aid stringlengths 9 15 | mid stringlengths 7 10 | abstract stringlengths 78 2.56k | related_work stringlengths 92 1.77k | ref_abstract dict |
|---|---|---|---|---|
1904.06109 | 2938008799 | In recent decades, 3D morphable model (3DMM) has been commonly used in image-based photorealistic 3D face reconstruction. However, face images are often corrupted by serious occlusion by non-face objects including eyeglasses, masks, and hands. Such objects block the correct capture of landmarks and shading information. Therefore, the reconstructed 3D face model is hardly reusable. In this paper, a novel method is proposed to restore de-occluded face images based on inverse use of 3DMM and generative adversarial network. We utilize the 3DMM prior to the proposed adversarial network and combine a global and local adversarial convolutional neural network to learn face de-occlusion model. The 3DMM serves not only as geometric prior but also proposes the face region for the local discriminator. Experiment results confirm the effectiveness and robustness of the proposed algorithm in removing challenging types of occlusions with various head poses and illumination. Furthermore, the proposed method reconstructs the correct 3D face model with de-occluded textures. | * -4mm Bernhard @cite_15 proposed an occlusion-aware face modeling method in which they incorporated 3DMM as appearance prior in a RANSAC-like algorithm. In this method, the input image is segmented into face and non-face regions and the illumination is estimated using the face region only. However, this method is not robust if the occlusions, e.g., hands, have a similar color appearance to the face. A recent face alignment technique shows the robustness in occluded face images @cite_2 @cite_35 . Thus, this technique can be used to fit 3DMM and produce a 3D face model with a few details. Although their goal is to robustly find the head pose, de-occlusion in not within the scope of their work. Tran @cite_1 is the first to address the problem of detailed face reconstruction from occluded images by filling in the corrupted region of the bump map using a similar patch in a reference dataset. Although this method can generate a complete representation of face details, the de-occluded face image is not reconstructed @cite_1 . | {
"cite_N": [
"@cite_35",
"@cite_15",
"@cite_1",
"@cite_2"
],
"mid": [
"",
"2786166435",
"2795709097",
"2605105738"
],
"abstract": [
"",
"Faces in natural images are often occluded by a variety of objects. We propose a fully automated, probabilistic and occlusion-aware 3D morphable face model adaptation framework following an analysis-by-synthesis setup. The key idea is to segment the image into regions explained by separate models. Our framework includes a 3D morphable face model, a prototype-based beard model and a simple model for occlusions and background regions. The segmentation and all the model parameters have to be inferred from the single target image. Face model adaptation and segmentation are solved jointly using an expectation–maximization-like procedure. During the E-step, we update the segmentation and in the M-step the face model parameters are updated. For face model adaptation we apply a stochastic sampling strategy based on the Metropolis–Hastings algorithm. For segmentation, we apply loopy belief propagation for inference in a Markov random field. Illumination estimation is critical for occlusion handling. Our combined segmentation and model adaptation needs a proper initialization of the illumination parameters. We propose a RANSAC-based robust illumination estimation technique. By applying this method to a large face image database we obtain a first empirical distribution of real-world illumination conditions. The obtained empirical distribution is made publicly available and can be used as prior in probabilistic frameworks, for regularization or to synthesize data for deep learning methods.",
"Existing single view, 3D face reconstruction methods can produce beautifully detailed 3D results, but typically only for near frontal, unobstructed viewpoints. We describe a system designed to provide detailed 3D reconstructions of faces viewed under extreme conditions, out of plane rotations, and occlusions. Motivated by the concept of bump mapping, we propose a layered approach which decouples estimation of a global shape from its mid-level details (e.g., wrinkles). We estimate a coarse 3D face shape which acts as a foundation and then separately layer this foundation with details represented by a bump map. We show how a deep convolutional encoder-decoder can be used to estimate such bump maps. We further show how this approach naturally extends to generate plausible details for occluded facial regions. We test our approach and its components extensively, quantitatively demonstrating the invariance of our estimated facial details. We further provide numerous qualitative examples showing that our method produces detailed 3D face shapes in viewing conditions where existing state of the art often break down.",
"This paper investigates how far a very deep neural network is from attaining close to saturating performance on existing 2D and 3D face alignment datasets. To this end, we make the following 5 contributions: (a) we construct, for the first time, a very strong baseline by combining a state-of-the-art architecture for landmark localization with a state-of-the-art residual block, train it on a very large yet synthetically expanded 2D facial landmark dataset and finally evaluate it on all other 2D facial landmark datasets. (b)We create a guided by 2D landmarks network which converts 2D landmark annotations to 3D and unifies all existing datasets, leading to the creation of LS3D-W, the largest and most challenging 3D facial landmark dataset to date ( 230,000 images). (c) Following that, we train a neural network for 3D face alignment and evaluate it on the newly introduced LS3D-W. (d) We further look into the effect of all “traditional” factors affecting face alignment performance like large pose, initialization and resolution, and introduce a “new” one, namely the size of the network. (e) We show that both 2D and 3D face alignment networks achieve performance of remarkable accuracy which is probably close to saturating the datasets used. Training and testing code as well as the dataset can be downloaded from https: www.adrianbulat.com face-alignment"
]
} |
1904.06109 | 2938008799 | In recent decades, 3D morphable model (3DMM) has been commonly used in image-based photorealistic 3D face reconstruction. However, face images are often corrupted by serious occlusion by non-face objects including eyeglasses, masks, and hands. Such objects block the correct capture of landmarks and shading information. Therefore, the reconstructed 3D face model is hardly reusable. In this paper, a novel method is proposed to restore de-occluded face images based on inverse use of 3DMM and generative adversarial network. We utilize the 3DMM prior to the proposed adversarial network and combine a global and local adversarial convolutional neural network to learn face de-occlusion model. The 3DMM serves not only as geometric prior but also proposes the face region for the local discriminator. Experiment results confirm the effectiveness and robustness of the proposed algorithm in removing challenging types of occlusions with various head poses and illumination. Furthermore, the proposed method reconstructs the correct 3D face model with de-occluded textures. | * -4mm GAN @cite_20 utilizes min–max optimization over the generator and discriminator and shows significant improvement in face synthesis applications, such as face attribute editing @cite_19 , face frontalization @cite_10 @cite_32 , and face completion @cite_5 @cite_23 . However, no previous study has been conducted on using GAN for de-occlusion on challenging faces. | {
"cite_N": [
"@cite_32",
"@cite_19",
"@cite_23",
"@cite_5",
"@cite_10",
"@cite_20"
],
"mid": [
"",
"2579578355",
"",
"2772024431",
"2963100452",
"2099471712"
],
"abstract": [
"",
"Face attributes are interesting due to their detailed description of human faces. Unlike prior researches working on attribute prediction, we address an inverse and more challenging problem called face attribute manipulation which aims at modifying a face image according to a given attribute value. Instead of manipulating the whole image, we propose to learn the corresponding residual image defined as the difference between images before and after the manipulation. In this way, the manipulation can be operated efficiently with modest pixel modification. The framework of our approach is based on the Generative Adversarial Network. It consists of two image transformation networks and a discriminative network. The transformation networks are responsible for the attribute manipulation and its dual operation and the discriminative network is used to distinguish the generated images from real images. We also apply dual learning to allow transformation networks to learn from each other. Experiments show that residual images can be effectively learned and used for attribute manipulations. The generated images remain most of the details in attribute-irrelevant areas.",
"",
"Recently proposed robust 3D face alignment methods establish either dense or sparse correspondence between a 3D face model and a 2D facial image. The use of these methods presents new challenges as well as opportunities for facial texture analysis. In particular, by sampling the image using the fitted model, a facial UV can be created. Unfortunately, due to self-occlusion, such a UV map is always incomplete. In this paper, we propose a framework for training Deep Convolutional Neural Network (DCNN) to complete the facial UV map extracted from in-the-wild images. To this end, we first gather complete UV maps by fitting a 3D Morphable Model (3DMM) to various multiview image and video datasets, as well as leveraging on a new 3D dataset with over 3,000 identities. Second, we devise a meticulously designed architecture that combines local and global adversarial DCNNs to learn an identity-preserving facial UV completion model. We demonstrate that by attaching the completed UV to the fitted mesh and generating instances of arbitrary poses, we can increase pose variations for training deep face recognition verification models, and minimise pose discrepancy during testing, which lead to better performance. Experiments on both controlled and in-the-wild UV datasets prove the effectiveness of our adversarial UV completion model. We achieve state-of-the-art verification accuracy, 94.05 , under the CFP frontal-profile protocol only by combining pose augmentation during training and pose discrepancy reduction during testing. We will release the first in-the-wild UV dataset (we refer as WildUV) that comprises of complete facial UV maps from 1,892 identities for research purposes.",
"Despite recent advances in face recognition using deep learning, severe accuracy drops are observed for large pose variations in unconstrained environments. Learning pose-invariant features is one solution, but needs expensively labeled large-scale data and carefully designed feature learning algorithms. In this work, we focus on frontalizing faces in the wild under various head poses, including extreme profile view's. We propose a novel deep 3D Morphable Model (3DMM) conditioned Face Frontalization Generative Adversarial Network (GAN), termed as FF-GAN, to generate neutral head pose face images. Our framework differs from both traditional GANs and 3DMM based modeling. Incorporating 3DMM into the GAN structure provides shape and appearance priors for fast convergence with less training data, while also supporting end-to-end training. The 3DMM-conditioned GAN employs not only the discriminator and generator loss but also a new masked symmetry loss to retain visual quality under occlusions, besides an identity loss to recover high frequency information. Experiments on face recognition, landmark localization and 3D reconstruction consistently show the advantage of our frontalization method on faces in the wild datasets. 1",
"We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a mistake. This framework corresponds to a minimax two-player game. In the space of arbitrary functions G and D, a unique solution exists, with G recovering the training data distribution and D equal to ½ everywhere. In the case where G and D are defined by multilayer perceptrons, the entire system can be trained with backpropagation. There is no need for any Markov chains or unrolled approximate inference networks during either training or generation of samples. Experiments demonstrate the potential of the framework through qualitative and quantitative evaluation of the generated samples."
]
} |
1904.05998 | 2936827896 | Information, ideas, and diseases, or more generally, contagions, spread over space and time through individual transmissions via social networks, as well as through external sources. A detailed picture of any diffusion process can be achieved only when both a good network structure and individual diffusion pathways are obtained. The advent of rich social, media and locational data allows us to study and model this diffusion process in more detail than previously possible. Nevertheless, how information, ideas or diseases are propagated through the network as an overall process is difficult to trace. This propagation is continuous over space and time, where individual transmissions occur at different rates via complex, latent connections. To tackle this challenge, a probabilistic spatiotemporal algorithm for network diffusion (STAND) is developed based on the survival model in this research. Both time and spatial distance are used as explanatory variables to simulate the diffusion process over two different network structures. The aim is to provide a more detailed measure of how different contagions are transmitted through various networks where nodes are geographic places at a large scale. | There is a large amount of recent research relating to diffusion over networks in geographic space via collecting and analyzing real-world data @cite_9 @cite_12 @cite_8 . When it comes to modeling the diffusion process at a large-scale, however, simulating the complete topology of spatiotemporal networks and multiple contagions spread over them becomes much more complex and challenging for two reasons: First, in many cases we can only observe the timing information of when nodes get infected @cite_0 . Second, even though great amounts of digital heterogeneous data are now available via the World Wide Web, locational information is extremely sparse, unstructured and often ambiguous. Developing a flexible model for deriving the network structure and cascade behaviors is key to uncovering the mechanism that governs spatiotemporal diffusion processes and their dynamics. | {
"cite_N": [
"@cite_0",
"@cite_9",
"@cite_12",
"@cite_8"
],
"mid": [
"2952347589",
"2342933002",
"2866540571",
"2913597007"
],
"abstract": [
"Time plays an essential role in the diffusion of information, influence and disease over networks. In many cases we only observe when a node copies information, makes a decision or becomes infected -- but the connectivity, transmission rates between nodes and transmission sources are unknown. Inferring the underlying dynamics is of outstanding interest since it enables forecasting, influencing and retarding infections, broadly construed. To this end, we model diffusion processes as discrete networks of continuous temporal processes occurring at different rates. Given cascade data -- observed infection times of nodes -- we infer the edges of the global diffusion network and estimate the transmission rates of each edge that best explain the observed data. The optimization problem is convex. The model naturally (without heuristics) imposes sparse solutions and requires no parameter tuning. The problem decouples into a collection of independent smaller problems, thus scaling easily to networks on the order of hundreds of thousands of nodes. Experiments on real and synthetic data show that our algorithm both recovers the edges of diffusion networks and accurately estimates their transmission rates from cascade data.",
"Abstract Social media data are increasingly being used for enhancing situational awareness and assisting disaster management. We analyzed the wildfire-related Twitter activities in terms of their attributes pertinent to space, time, content, and network, so as to gain insights into the usefulness of social media data in revealing situational awareness. Findings show that social media data can characterize the wildfire across space and over time, and thus are applicable to provide useful information on disaster situations. Second, people have strong geographical awareness during wildfire hazards and are interested in communicating situational updates related to wildfire damage (e.g., containment percentage and burned acres), wildfire response (e.g., evacuation), and appreciation to firefighters. Third, news media and local authorities are opinion leaders and play a dominant role in the wildfire retweet network.",
"To address the effects of increasing homeless populations, planners must understand the size and distribution of their homeless populations, as well as how information and resources are diffused throughout homeless communities. Currently, there is limited publicly available information on the homeless population, e.g. the estimates of the homeless, gathered annually by the US Housing & Urban Development point in time survey. While it is theorized in the literature that the networks of homeless individuals provide access to important information for planners in areas such as health (e.g. needle exchanges) or access (e.g. information diffusion about the location of new shelters), it is almost never measured, and if measured, only at a very small scale. This research addresses the question of how planners can leverage publicly available data on the homeless to better understand their own homeless networks (e.g. relations among the homeless themselves) in a cost-effective and reliable way. To this end, we pro...",
"Abstract A two-layer susceptible-infected-recovered unaware-aware (SIR-UA) epidemic model is presented to analyze the effect of different heterogeneous networks in a population. Random, scale-free, and small-world network topologies are tested to investigate the impact of awareness on the spread of epidemics in a two-layer network with diverse combinations of degree and structure. Susceptible and infected (both unaware and aware) individuals are associated with their neighboring nodes in a social network structure with various degree distributions. In the two-layer SIR-UA epidemic model, a virtual network represents the connections that spread information, while a physical network represents the physical social interactions that spread diseases. We test various combinations of network structures in virtual or physical networks, to understand the impact of information diffusion on the spread of epidemics in a heterogeneous network structure. Then, the effects of awareness on the spread of a disease are discussed. Finally, phase diagrams are illustrated to reveal the final regions covered by an epidemic with various network parameters. We find that a disease spreads less if the virtual social network is more connected than the network of physical connections."
]
} |
1904.05998 | 2936827896 | Information, ideas, and diseases, or more generally, contagions, spread over space and time through individual transmissions via social networks, as well as through external sources. A detailed picture of any diffusion process can be achieved only when both a good network structure and individual diffusion pathways are obtained. The advent of rich social, media and locational data allows us to study and model this diffusion process in more detail than previously possible. Nevertheless, how information, ideas or diseases are propagated through the network as an overall process is difficult to trace. This propagation is continuous over space and time, where individual transmissions occur at different rates via complex, latent connections. To tackle this challenge, a probabilistic spatiotemporal algorithm for network diffusion (STAND) is developed based on the survival model in this research. Both time and spatial distance are used as explanatory variables to simulate the diffusion process over two different network structures. The aim is to provide a more detailed measure of how different contagions are transmitted through various networks where nodes are geographic places at a large scale. | While, most existing simulation models of network diffusion are based on assumptions of agent homogeneity or how agents interact with each other for spreading contagions, none have integrated space and time together to account for diffusion probability over spatiotemporal networks at large-scale. Our goal is to fill this gap. We develop an exponential diffusion probabilistic algorithm inspired by NETINF @cite_0 that integrates geographic distance, node infection time and transmission speed together. The reasons to choose the network diffusion model underlying NETINF as a starting point for STAND are: 1) it's a continuous-time model without any assumptions about the individual interactions, 2) it is amenable to inference estimation for different datasets, and 3) it has a flexible structure in terms of adding features like geo-distance. Other time distributions (e.g., log-normal, Gamma, etc.) can be considered in the future research as well. | {
"cite_N": [
"@cite_0"
],
"mid": [
"2952347589"
],
"abstract": [
"Time plays an essential role in the diffusion of information, influence and disease over networks. In many cases we only observe when a node copies information, makes a decision or becomes infected -- but the connectivity, transmission rates between nodes and transmission sources are unknown. Inferring the underlying dynamics is of outstanding interest since it enables forecasting, influencing and retarding infections, broadly construed. To this end, we model diffusion processes as discrete networks of continuous temporal processes occurring at different rates. Given cascade data -- observed infection times of nodes -- we infer the edges of the global diffusion network and estimate the transmission rates of each edge that best explain the observed data. The optimization problem is convex. The model naturally (without heuristics) imposes sparse solutions and requires no parameter tuning. The problem decouples into a collection of independent smaller problems, thus scaling easily to networks on the order of hundreds of thousands of nodes. Experiments on real and synthetic data show that our algorithm both recovers the edges of diffusion networks and accurately estimates their transmission rates from cascade data."
]
} |
1904.05729 | 2939256164 | As a sub-domain of text-to-image synthesis, text-to-face generation has huge potentials in public safety domain. With lack of dataset, there are almost no related research focusing on text-to-face synthesis. In this paper, we propose a fully-trained Generative Adversarial Network (FTGAN) that trains the text encoder and image decoder at the same time for fine-grained text-to-face generation. With a novel fully-trained generative network, FTGAN can synthesize higher-quality images and urge the outputs of the FTGAN are more relevant to the input sentences. In addition, we build a dataset called SCU-Text2face for text-to-face synthesis. Through extensive experiments, the FTGAN shows its superiority in boosting both generated images' quality and similarity to the input descriptions. The proposed FTGAN outperforms the previous state of the art, boosting the best reported Inception Score to 4.63 on the CUB dataset. On SCU-text2face, the face images generated by our proposed FTGAN just based on the input descriptions is of average 59 similarity to the ground-truth, which set a baseline for text-to-face synthesis. | Despite there are kinds of networks for text-to-image synthesis, they are mostly based on encoder-decoder framework and conditional GAN @cite_2 . This encoder-decoder framework inludes text-encoder and image-decoder. The text-encoder turn input descriptions to semantic vectors and the image-decoder turn the encoded semantic vectors to natural images. There are two main targets for text-to-image synthesis: to generate high-quality images and generate images matching the given descriptions. All the developments of text-to-images synthesis are based on this two targets. | {
"cite_N": [
"@cite_2"
],
"mid": [
"2125389028"
],
"abstract": [
"Generative Adversarial Nets [8] were recently introduced as a novel way to train generative models. In this work we introduce the conditional version of generative adversarial nets, which can be constructed by simply feeding the data, y, we wish to condition on to both the generator and discriminator. We show that this model can generate MNIST digits conditioned on class labels. We also illustrate how this model could be used to learn a multi-modal model, and provide preliminary examples of an application to image tagging in which we demonstrate how this approach can generate descriptive tags which are not part of training labels."
]
} |
1904.05729 | 2939256164 | As a sub-domain of text-to-image synthesis, text-to-face generation has huge potentials in public safety domain. With lack of dataset, there are almost no related research focusing on text-to-face synthesis. In this paper, we propose a fully-trained Generative Adversarial Network (FTGAN) that trains the text encoder and image decoder at the same time for fine-grained text-to-face generation. With a novel fully-trained generative network, FTGAN can synthesize higher-quality images and urge the outputs of the FTGAN are more relevant to the input sentences. In addition, we build a dataset called SCU-Text2face for text-to-face synthesis. Through extensive experiments, the FTGAN shows its superiority in boosting both generated images' quality and similarity to the input descriptions. The proposed FTGAN outperforms the previous state of the art, boosting the best reported Inception Score to 4.63 on the CUB dataset. On SCU-text2face, the face images generated by our proposed FTGAN just based on the input descriptions is of average 59 similarity to the ground-truth, which set a baseline for text-to-face synthesis. | The early research for text-to-image synthesis are mainly focusing on improving the quality of generated images. The task text-to-image was first presented in 2016, Reed al presented this novel task and developed two end-to-end networks based on conditional GAN to accomplish it @cite_26 . Reed utilized a pretrained Char-CNN-RNN network for text encoding and built a network similar to DCGAN @cite_24 as image decoder to generate natural images from vector. Then many researchers made some progresses based on his work @cite_14 . One of the most influential research is made by Zhang al, they proposed a 2-stages network StackGAN to solve this task, which could generate high-quality images and improved the Inception Score obviously @cite_4 . This network is also inherited by later research @cite_6 @cite_29 @cite_18 @cite_31 . | {
"cite_N": [
"@cite_18",
"@cite_14",
"@cite_26",
"@cite_4",
"@cite_29",
"@cite_6",
"@cite_24",
"@cite_31"
],
"mid": [
"2963413689",
"2964313012",
"2405756170",
"2964024144",
"2963966654",
"2963163163",
"2963684088",
""
],
"abstract": [
"This paper presents a novel method to deal with the challenging task of generating photographic images conditioned on semantic image descriptions. Our method introduces accompanying hierarchical-nested adversarial objectives inside the network hierarchies, which regularize mid-level representations and assist generator training to capture the complex image statistics. We present an extensile single-stream generator architecture to better adapt the jointed discriminators and push generated images up to high resolutions. We adopt a multi-purpose adversarial loss to encourage more effective image and text information usage in order to improve the semantic consistency and image fidelity simultaneously. Furthermore, we introduce a new visual-semantic similarity measure to evaluate the semantic consistency of generated images. With extensive experimental validation on three public datasets, our method significantly improves previous state of the arts on all datasets over different evaluation metrics.",
"In this paper, we propose a way of synthesizing realistic images directly with natural language description, which has many useful applications, e.g. intelligent image manipulation. We attempt to accomplish such synthesis: given a source image and a target text description, our model synthesizes images to meet two requirements: 1) being realistic while matching the target text description; 2) maintaining other image features that are irrelevant to the text description. The model should be able to disentangle the semantic information from the two modalities (image and text), and generate new images from the combined semantics. To achieve this, we proposed an end-to-end neural architecture that leverages adversarial learning to automatically learn implicit loss functions, which are optimized to fulfill the aforementioned two requirements. We have evaluated our model by conducting experiments on Caltech-200 bird dataset and Oxford-102 flower dataset, and have demonstrated that our model is capable of synthesizing realistic images that match the given descriptions, while still maintain other features of original images.",
"Automatic synthesis of realistic images from text would be interesting and useful, but current AI systems are still far from this goal. However, in recent years generic and powerful recurrent neural network architectures have been developed to learn discriminative text feature representations. Meanwhile, deep convolutional generative adversarial networks (GANs) have begun to generate highly compelling images of specific categories, such as faces, album covers, and room interiors. In this work, we develop a novel deep architecture and GAN formulation to effectively bridge these advances in text and image modeling, translating visual concepts from characters to pixels. We demonstrate the capability of our model to generate plausible images of birds and flowers from detailed text descriptions.",
"Synthesizing high-quality images from text descriptions is a challenging problem in computer vision and has many practical applications. Samples generated by existing textto- image approaches can roughly reflect the meaning of the given descriptions, but they fail to contain necessary details and vivid object parts. In this paper, we propose Stacked Generative Adversarial Networks (StackGAN) to generate 256.256 photo-realistic images conditioned on text descriptions. We decompose the hard problem into more manageable sub-problems through a sketch-refinement process. The Stage-I GAN sketches the primitive shape and colors of the object based on the given text description, yielding Stage-I low-resolution images. The Stage-II GAN takes Stage-I results and text descriptions as inputs, and generates high-resolution images with photo-realistic details. It is able to rectify defects in Stage-I results and add compelling details with the refinement process. To improve the diversity of the synthesized images and stabilize the training of the conditional-GAN, we introduce a novel Conditioning Augmentation technique that encourages smoothness in the latent conditioning manifold. Extensive experiments and comparisons with state-of-the-arts on benchmark datasets demonstrate that the proposed method achieves significant improvements on generating photo-realistic images conditioned on text descriptions.",
"In this paper, we propose an Attentional Generative Adversarial Network (AttnGAN) that allows attention-driven, multi-stage refinement for fine-grained text-to-image generation. With a novel attentional generative network, the AttnGAN can synthesize fine-grained details at different sub-regions of the image by paying attentions to the relevant words in the natural language description. In addition, a deep attentional multimodal similarity model is proposed to compute a fine-grained image-text matching loss for training the generator. The proposed AttnGAN significantly outperforms the previous state of the art, boosting the best reported inception score by 14.14 on the CUB dataset and 170.25 on the more challenging COCO dataset. A detailed analysis is also performed by visualizing the attention layers of the AttnGAN. It for the first time shows that the layered attentional GAN is able to automatically select the condition at the word level for generating different parts of the image.",
"Although Generative Adversarial Networks (GANs) have shown remarkable success in various tasks, they still face challenges in generating high quality images. In this paper, we propose Stacked Generative Adversarial Networks (StackGANs) aimed at generating high-resolution photo-realistic images. First, we propose a two-stage generative adversarial network architecture, StackGAN-v1, for text-to-image synthesis. The Stage-I GAN sketches the primitive shape and colors of a scene based on a given text description, yielding low-resolution images. The Stage-II GAN takes Stage-I results and the text description as inputs, and generates high-resolution images with photo-realistic details. Second, an advanced multi-stage generative adversarial network architecture, StackGAN-v2, is proposed for both conditional and unconditional generative tasks. Our StackGAN-v2 consists of multiple generators and multiple discriminators arranged in a tree-like structure; images at multiple scales corresponding to the same scene are generated from different branches of the tree. StackGAN-v2 shows more stable training behavior than StackGAN-v1 by jointly approximating multiple distributions. Extensive experiments demonstrate that the proposed stacked generative adversarial networks significantly outperform other state-of-the-art methods in generating photo-realistic images.",
"Abstract: In recent years, supervised learning with convolutional networks (CNNs) has seen huge adoption in computer vision applications. Comparatively, unsupervised learning with CNNs has received less attention. In this work we hope to help bridge the gap between the success of CNNs for supervised learning and unsupervised learning. We introduce a class of CNNs called deep convolutional generative adversarial networks (DCGANs), that have certain architectural constraints, and demonstrate that they are a strong candidate for unsupervised learning. Training on various image datasets, we show convincing evidence that our deep convolutional adversarial pair learns a hierarchy of representations from object parts to scenes in both the generator and discriminator. Additionally, we use the learned features for novel tasks - demonstrating their applicability as general image representations.",
""
]
} |
1904.05729 | 2939256164 | As a sub-domain of text-to-image synthesis, text-to-face generation has huge potentials in public safety domain. With lack of dataset, there are almost no related research focusing on text-to-face synthesis. In this paper, we propose a fully-trained Generative Adversarial Network (FTGAN) that trains the text encoder and image decoder at the same time for fine-grained text-to-face generation. With a novel fully-trained generative network, FTGAN can synthesize higher-quality images and urge the outputs of the FTGAN are more relevant to the input sentences. In addition, we build a dataset called SCU-Text2face for text-to-face synthesis. Through extensive experiments, the FTGAN shows its superiority in boosting both generated images' quality and similarity to the input descriptions. The proposed FTGAN outperforms the previous state of the art, boosting the best reported Inception Score to 4.63 on the CUB dataset. On SCU-text2face, the face images generated by our proposed FTGAN just based on the input descriptions is of average 59 similarity to the ground-truth, which set a baseline for text-to-face synthesis. | Since GAN was proposed by Goodfellow in 2014 @cite_12 , image synthesis has been a hot topic in deep learning. Because there are two large scale public dataset: CelebA and LFW @cite_0 , face synthesis is also a popular research domain. Almost most of the state-of-art networks will examplify their model's superiority on face synthesis, including networks based on GAN and networks based both on conditional GAN(such as DCGAN @cite_24 , CycleGAN @cite_34 , ProGAN @cite_16 , BigGAN @cite_13 , StyleGAN @cite_5 , Stargan @cite_32 and so on). With the development of those networks, the quality of generated face images are becoming better and better. Now some networks could even generate 1024 @math 1024 face images, much larger than the original images resolution of the face dataset. Those models aim to learn a mapping from noise vector which conforms to Normal distribution to natural face images. But they can't control the network to generate a precise face image which they want. | {
"cite_N": [
"@cite_32",
"@cite_34",
"@cite_0",
"@cite_24",
"@cite_5",
"@cite_16",
"@cite_13",
"@cite_12"
],
"mid": [
"2768626898",
"2962793481",
"1782590233",
"2963684088",
"2904367110",
"2962760235",
"2952716587",
"2099471712"
],
"abstract": [
"Recent studies have shown remarkable success in image-to-image translation for two domains. However, existing approaches have limited scalability and robustness in handling more than two domains, since different models should be built independently for every pair of image domains. To address this limitation, we propose StarGAN, a novel and scalable approach that can perform image-to-image translations for multiple domains using only a single model. Such a unified model architecture of StarGAN allows simultaneous training of multiple datasets with different domains within a single network. This leads to StarGAN's superior quality of translated images compared to existing models as well as the novel capability of flexibly translating an input image to any desired target domain. We empirically demonstrate the effectiveness of our approach on a facial attribute transfer and a facial expression synthesis tasks.",
"Image-to-image translation is a class of vision and graphics problems where the goal is to learn the mapping between an input image and an output image using a training set of aligned image pairs. However, for many tasks, paired training data will not be available. We present an approach for learning to translate an image from a source domain X to a target domain Y in the absence of paired examples. Our goal is to learn a mapping G : X → Y such that the distribution of images from G(X) is indistinguishable from the distribution Y using an adversarial loss. Because this mapping is highly under-constrained, we couple it with an inverse mapping F : Y → X and introduce a cycle consistency loss to push F(G(X)) ≈ X (and vice versa). Qualitative results are presented on several tasks where paired training data does not exist, including collection style transfer, object transfiguration, season transfer, photo enhancement, etc. Quantitative comparisons against several prior methods demonstrate the superiority of our approach.",
"Most face databases have been created under controlled conditions to facilitate the study of specific parameters on the face recognition problem. These parameters include such variables as position, pose, lighting, background, camera quality, and gender. While there are many applications for face recognition technology in which one can control the parameters of image acquisition, there are also many applications in which the practitioner has little or no control over such parameters. This database, Labeled Faces in the Wild, is provided as an aid in studying the latter, unconstrained, recognition problem. The database contains labeled face photographs spanning the range of conditions typically encountered in everyday life. The database exhibits “natural” variability in factors such as pose, lighting, race, accessories, occlusions, and background. In addition to describing the details of the database, we provide specific experimental paradigms for which the database is suitable. This is done in an effort to make research performed with the database as consistent and comparable as possible. We provide baseline results, including results of a state of the art face recognition system combined with a face alignment system. To facilitate experimentation on the database, we provide several parallel databases, including an aligned version.",
"Abstract: In recent years, supervised learning with convolutional networks (CNNs) has seen huge adoption in computer vision applications. Comparatively, unsupervised learning with CNNs has received less attention. In this work we hope to help bridge the gap between the success of CNNs for supervised learning and unsupervised learning. We introduce a class of CNNs called deep convolutional generative adversarial networks (DCGANs), that have certain architectural constraints, and demonstrate that they are a strong candidate for unsupervised learning. Training on various image datasets, we show convincing evidence that our deep convolutional adversarial pair learns a hierarchy of representations from object parts to scenes in both the generator and discriminator. Additionally, we use the learned features for novel tasks - demonstrating their applicability as general image representations.",
"We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces.",
"We describe a new training methodology for generative adversarial networks. The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. This both speeds the training up and greatly stabilizes it, allowing us to produce images of unprecedented quality, e.g., CelebA images at 1024^2. We also propose a simple way to increase the variation in generated images, and achieve a record inception score of 8.80 in unsupervised CIFAR10. Additionally, we describe several implementation details that are important for discouraging unhealthy competition between the generator and discriminator. Finally, we suggest a new metric for evaluating GAN results, both in terms of image quality and variation. As an additional contribution, we construct a higher-quality version of the CelebA dataset.",
"Despite recent progress in generative image modeling, successfully generating high-resolution, diverse samples from complex datasets such as ImageNet remains an elusive goal. To this end, we train Generative Adversarial Networks at the largest scale yet attempted, and study the instabilities specific to such scale. We find that applying orthogonal regularization to the generator renders it amenable to a simple \"truncation trick,\" allowing fine control over the trade-off between sample fidelity and variety by reducing the variance of the Generator's input. Our modifications lead to models which set the new state of the art in class-conditional image synthesis. When trained on ImageNet at 128x128 resolution, our models (BigGANs) achieve an Inception Score (IS) of 166.5 and Frechet Inception Distance (FID) of 7.4, improving over the previous best IS of 52.52 and FID of 18.6.",
"We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a mistake. This framework corresponds to a minimax two-player game. In the space of arbitrary functions G and D, a unique solution exists, with G recovering the training data distribution and D equal to ½ everywhere. In the case where G and D are defined by multilayer perceptrons, the entire system can be trained with backpropagation. There is no need for any Markov chains or unrolled approximate inference networks during either training or generation of samples. Experiments demonstrate the potential of the framework through qualitative and quantitative evaluation of the generated samples."
]
} |
1904.05729 | 2939256164 | As a sub-domain of text-to-image synthesis, text-to-face generation has huge potentials in public safety domain. With lack of dataset, there are almost no related research focusing on text-to-face synthesis. In this paper, we propose a fully-trained Generative Adversarial Network (FTGAN) that trains the text encoder and image decoder at the same time for fine-grained text-to-face generation. With a novel fully-trained generative network, FTGAN can synthesize higher-quality images and urge the outputs of the FTGAN are more relevant to the input sentences. In addition, we build a dataset called SCU-Text2face for text-to-face synthesis. Through extensive experiments, the FTGAN shows its superiority in boosting both generated images' quality and similarity to the input descriptions. The proposed FTGAN outperforms the previous state of the art, boosting the best reported Inception Score to 4.63 on the CUB dataset. On SCU-text2face, the face images generated by our proposed FTGAN just based on the input descriptions is of average 59 similarity to the ground-truth, which set a baseline for text-to-face synthesis. | To tackle this issue, with conditional GAN, face synthesis have derived many interesting applications about face, such as translating edges to natural face images @cite_17 , exchanging the attributes of two face images @cite_33 , generating a positive face from the side face @cite_28 , generating a full face from eyes' region only @cite_3 , from face attributes to sketches to natural face images synthesis @cite_37 ,face inpainting @cite_27 and so on. Those networks try to control the synthesized face images by adding a condition vector, could generate face images which meet the needs of different situations. Text-to-face synthesis is similar to those tasks, which utilize the input descriptions as the control condition. | {
"cite_N": [
"@cite_37",
"@cite_33",
"@cite_28",
"@cite_3",
"@cite_27",
"@cite_17"
],
"mid": [
"2781490755",
"2964339532",
"2964337551",
"2792829392",
"2963540914",
"2963800363"
],
"abstract": [
"Automatic synthesis of faces from visual attributes is an important problem in computer vision and has wide applications in law enforcement and entertainment. With the advent of deep generative convolutional neural networks (CNNs), attempts have been made to synthesize face images from attributes and text descriptions. In this paper, we take a different approach, where we formulate the original problem as a stage-wise learning problem. We first synthesize the facial sketch corresponding to the visual attributes and then we reconstruct the face image based on the synthesized sketch. The proposed Attribute2Sketch2Face framework, which is based on a combination of deep Conditional Variational Autoencoder (CVAE) and Generative Adversarial Networks (GANs), consists of three stages: (1) Synthesis of facial sketch from attributes using a CVAE architecture, (2) Enhancement of coarse sketches to produce sharper sketches using a GAN-based framework, and (3) Synthesis of face from sketch using another GAN-based network. Extensive experiments and comparison with recent methods are performed to verify the effectiveness of the proposed attribute-based three stage face synthesis method.",
"We propose a framework based on Generative Adversarial Networks to disentangle the identity and attributes of faces, such that we can conveniently recombine different identities and attributes for identity preserving face synthesis in open domains. Previous identity preserving face synthesis processes are largely confined to synthesizing faces with known identities that are already in the training dataset. To synthesize a face with identity outside the training dataset, our framework requires one input image of that subject to produce an identity vector, and any other input face image to extract an attribute vector capturing, e.g., pose, emotion, illumination, and even the background. We then recombine the identity vector and the attribute vector to synthesize a new face of the subject with the extracted attribute. Our proposed framework does not need to annotate the attributes of faces in any way. It is trained with an asymmetric loss function to better preserve the identity and stabilize the training process. It can also effectively leverage large amounts of unlabeled training face images to further improve the fidelity of the synthesized faces for subjects that are not presented in the labeled training face dataset. Our experiments demonstrate the efficacy of the proposed framework. We also present its usage in a much broader set of applications including face frontalization, face attribute morphing, and face adversarial example detection.",
"Photorealistic frontal view synthesis from a single face image has a wide range of applications in the field of face recognition. Although data-driven deep learning methods have been proposed to address this problem by seeking solutions from ample face data, this problem is still challenging because it is intrinsically ill-posed. This paper proposes a Two-Pathway Generative Adversarial Network (TP-GAN) for photorealistic frontal view synthesis by simultaneously perceiving global structures and local details. Four landmark located patch networks are proposed to attend to local textures in addition to the commonly used global encoderdecoder network. Except for the novel architecture, we make this ill-posed problem well constrained by introducing a combination of adversarial loss, symmetry loss and identity preserving loss. The combined loss function leverages both frontal face distribution and pre-trained discriminative deep face models to guide an identity preserving inference of frontal views from profiles. Different from previous deep learning methods that mainly rely on intermediate features for recognition, our method directly leverages the synthesized identity preserving image for downstream tasks like face recognition and attribution estimation. Experimental results demonstrate that our method not only presents compelling perceptual results but also outperforms state-of-theart results on large pose face recognition.",
"With the popularity of surveillance cameras and the development of deep learning, significant progress has been made in the field of smart surveillance. Face recognition is one of the most important yet challenging tasks in human-centered smart surveillance, especially in public security, criminal investigation and anti-terrorism, and so on. Although, the state-of-the-art algorithms for face recognition have achieved dramatically improved results and have been widely applied in authentication scenario, the occlusion problem on face is still one of the critical issues for personal identification in smart surveillance, especially in the occasion of terrorist searching and identification. To address this issue, this paper proposed a new approach for eyes-to-face synthesis and personal identification for human-centered smart surveillance. An end-to-end network based on conditional generative adversarial networks (GAN) is designed to generate the face information based only on the available data of eyes region. To obtain photorealistic faces and identity-preserving information, a synthesis loss function based on feature loss, GAN loss, and total variation loss is proposed to guide the training process. Both the subject and objective experimental results demonstrated that the proposed method can preserve the identity based on eyes-only data, and provide a potential solution for the identification of person even in the case of face occlusion.",
"Recent deep learning based approaches have shown promising results for the challenging task of inpainting large missing regions in an image. These methods can generate visually plausible image structures and textures, but often create distorted structures or blurry textures inconsistent with surrounding areas. This is mainly due to ineffectiveness of convolutional neural networks in explicitly borrowing or copying information from distant spatial locations. On the other hand, traditional texture and patch synthesis approaches are particularly suitable when it needs to borrow textures from the surrounding regions. Motivated by these observations, we propose a new deep generative model-based approach which can not only synthesize novel image structures but also explicitly utilize surrounding image features as references during network training to make better predictions. The model is a feedforward, fully convolutional neural network which can process images with multiple holes at arbitrary locations and with variable sizes during the test time. Experiments on multiple datasets including faces (CelebA, CelebA-HQ), textures (DTD) and natural images (ImageNet, Places2) demonstrate that our proposed approach generates higher-quality inpainting results than existing ones. Code, demo and models are available at: https: github.com JiahuiYu generative_inpainting.",
"We present a new method for synthesizing high-resolution photo-realistic images from semantic label maps using conditional generative adversarial networks (conditional GANs). Conditional GANs have enabled a variety of applications, but the results are often limited to low-resolution and still far from realistic. In this work, we generate 2048 A— 1024 visually appealing results with a novel adversarial loss, as well as new multi-scale generator and discriminator architectures. Furthermore, we extend our framework to interactive visual manipulation with two additional features. First, we incorporate object instance segmentation information, which enables object manipulations such as removing adding objects and changing the object category. Second, we propose a method to generate diverse results given the same input, allowing users to edit the object appearance interactively. Human opinion studies demonstrate that our method significantly outperforms existing methods, advancing both the quality and the resolution of deep image synthesis and editing."
]
} |
1904.05775 | 2937415931 | Liquid democracy is a collective decision making paradigm which lies between direct and representative democracy. One of its main features is that voters can delegate their votes in a transitive manner such that: A delegates to B and B delegates to C leads to A indirectly delegates to C. These delegations can be effectively empowered by implementing liquid democracy in a social network, so that voters can delegate their votes to any of their neighbors in the network. However, it is uncertain that such a delegation process will lead to a stable state where all voters are satisfied with the people representing them. We study the stability (w.r.t. voters preferences) of the delegation process in liquid democracy and model it as a game in which the players are the voters and the strategies are their possible delegations. We answer several questions on the equilibria of this process in any social network or in social networks that correspond to restricted types of graphs. We show that a Nash-equilibrium may not exist, and that it is even NP-complete to decide whether one exists or not. This holds even if the social network is a complete graph or a bounded degree graph. We further show that this existence problem is W[1]-hard w.r.t. the treewidth of the social network. Besides these hardness results, we demonstrate that an equilibrium always exists whatever the preferences of the voters iff the social network is a tree. We design a dynamic programming procedure to determine some desirable equilibria (e.g., minimizing the dissatisfaction of the voters) in polynomial time for tree social networks. Lastly, we study the convergence of delegation dynamics. Unfortunately, when an equilibrium exists, we show that a best response dynamics may not converge, even if the social network is a path or a complete graph. | Our approach is closest to this last work as we consider the same type of delegation games. However, our model of preferences over guru is more general as we assume that each voter has a preference order over her possible gurus. These preference orders may be dictated by competence levels and types of voters as in @cite_7 . However, we do not make such hypothesis as the criteria to choose a delegate are numerous: geographic locality, cultural, political or religious identity, et caetera. Considering this more general framework strongly modifies the resulting delegation games. | {
"cite_N": [
"@cite_7"
],
"mid": [
"2964156872"
],
"abstract": [
"Liquid democracy is a proxy voting method where proxies are delegable. We propose and study a game-theoretic model of liquid democracy to address the following question: when is it rational for a voter to delegate her vote? We study the existence of pure-strategy Nash equilibria in this model, and how group accuracy is affected by them. We complement these theoretical results by means of agent-based simulations to study the effects of delegations on group's accuracy on variously structured social networks."
]
} |
1904.05796 | 2952628729 | This paper presents our approach to develop a method for an unmanned ground vehicle (UGV) to perform inspection tasks in nuclear environments using rich information maps. To reduce inspectors' exposure to elevated radiation levels, an autonomous navigation framework for the UGV has been developed to perform routine inspections such as counting containers, recording their ID tags and performing gamma measurements on some of them. In order to achieve autonomy, a rich information map is generated which includes not only the 2D global cost map consisting of obstacle locations for path planning, but also the location and orientation information for the objects of interest from the inspector's perspective. The UGV's autonomy framework utilizes this information to prioritize locations to navigate to perform the inspections. In this paper, we present our method of generating this rich information map, originally developed to meet the requirements of the International Atomic Energy Agency (IAEA) Robotics Challenge. We demonstrate the performance of our method in a simulated testbed environment containing uranium hexafluoride (UF6) storage container mock ups. | In addition to generating a semantic map, the fusion of SLAM and object detection has many other applications. Object recognition can be enhanced by incorporating multiple viewpoint detection with the support from SLAM @cite_6 . For instance, @cite_12 uses object detection to achieve a robust SLAM in home environments by treating detected objects as landmarks. Object detection can also improve SLAM quality by detecting and removing occlusion such as walking people in the static environment @cite_10 . @cite_14 takes advantage of object detection to assist a blind person without generating a map. They detect structural objects such as stairways or doorways and use the detected results to navigate a blind person to a destination. | {
"cite_N": [
"@cite_14",
"@cite_10",
"@cite_12",
"@cite_6"
],
"mid": [
"2753027569",
"2767841593",
"2110684306",
"2963288928"
],
"abstract": [
"This paper presents a 3-D object recognition method and its implementation on a robotic navigation aid to allow real-time detection of indoor structural objects for the navigation of a blind person. The method segments a point cloud into numerous planar patches and extracts their inter-plane relationships (IPRs).Based on the existing IPRs of the object models, the method defines six high level features (HLFs) and determines the HLFs for each patch. A Gaussian-mixture-model-based plane classifier is then devised to classify each planar patch into one belonging to a particular object model. Finally, a recursive plane clustering procedure is used to cluster the classified planes into the model objects. As the proposed method uses geometric context to detect an object, it is robust to the object’s visual appearance change. As a result, it is ideal for detecting structural objects (e.g., stairways, doorways, and so on). In addition, it has high scalability and parallelism. The method is also capable of detecting some indoor non-structural objects. Experimental results demonstrate that the proposed method has a high success rate in object recognition.",
"We propose a visual SLAM (Simultaneous Localization And Mapping) system able to perform robustly in populated environments. The image stream from a moving RGB-D camera is the only input to the system. The computed map in real-time is composed of two layers: 1) The unpopulated geometrical layer, which describes the geometry of the bare scene as an occupancy grid where pieces of information corresponding to people have been removed; 2) A semantic human activity layer, which describes the trajectory of each person with respect to the unpopulated map, labelling an area as \"traversable\" or \"occupied\". Our proposal is to embed a real-time human tracker into the system. The purpose is twofold. First, to mask out of the rigid SLAM pipeline the image regions occupied by people, which boosts the robustness, the relocation, the accuracy and the reusability of the geometrical map in populated scenes. Secondly, to estimate the full trajectory of each detected person with respect to the scene map, irrespective of the location of the moving camera when the person was imaged. The proposal is tested with two popular visual SLAM systems, C2TAM and ORBSLAM2, proving its generality. The experiments process a benchmark of RGB-D sequences from camera onboard a mobile robot. They prove the robustness, accuracy and reuse capabilities of the two layer map for populated scenes.",
"Reliable data association is crucial to localization and map building for mobile robot applications. For that reason, many mobile robots tend to choose vision-based SLAM solutions. In this paper, a SLAM scheme based on visual object recognition, not just a scene matching, in home environment is proposed without using artificial landmarks. For the object-based SLAM, the following algorithms are suggested: 1) a novel local invariant feature extraction by combining advantages of multi-scale Harris corner as a detector and its SIFT descriptor for natural object recognition, 2) the RANSAC clustering for robust object recognition in the presence of outliers and 3) calculating accurate metric information for SLAM update. The proposed algorithms increase robustness by correct data association and accurate observation. Moreover, it also can be easily implemented real-time by reducing the number of representative landmarks, i.e. objects. The performance of the proposed algorithm was verified by experiments using EKF-SLAM with a stereo camera in home-like environments, and it showed that the final pose error was bounded after battery-run-out autonomous navigation for 50 minutes",
""
]
} |
1904.05548 | 2937725239 | We propose a novel model to address the task of Visual Dialog which exhibits complex dialog structures. To obtain a reasonable answer based on the current question and the dialog history, the underlying semantic dependencies between dialog entities are essential. In this paper, we explicitly formalize this task as inference in a graphical model with partially observed nodes and unknown graph structures (relations in dialog). The given dialog entities are viewed as the observed nodes. The answer to a given question is represented by a node with missing value. We first introduce an Expectation Maximization algorithm to infer both the underlying dialog structures and the missing node values (desired answers). Based on this, we proceed to propose a differentiable graph neural network (GNN) solution that approximates this process. Experiment results on the VisDial and VisDial-Q datasets show that our model outperforms comparative methods. It is also observed that our method can infer the underlying dialog structure for better dialog reasoning. | aims to annotate images with natural language at the scene level automatically, which has been a long-term active research area in computer vision community. Early work @cite_65 @cite_16 typically tackled this task as a retrieval problem, , finding the best fitting caption from a set of predetermined caption templates. Modern methods @cite_0 @cite_32 @cite_53 were mainly based on a CNN-RNN framework, where the RNN leverages the CNN-representation of an input image to output a word sequence as the caption. In this way, they were freed from dependence of the pre-defined, expression-limited caption candidate pool. After that, some methods @cite_62 @cite_56 @cite_30 tried to integrate the vanilla CNN-RNN architecture with neural attention mechanisms, like semantic attention @cite_30 , and bottom-up top-down attention @cite_56 , to name a few representative ones. Another popular trend @cite_46 @cite_11 @cite_13 @cite_54 @cite_36 @cite_55 @cite_17 in this area focuses on improving the discriminability of caption generations, such as stylized image captioning @cite_46 @cite_17 , personalized image captioning @cite_11 , and context-aware image captioning @cite_13 @cite_54 . | {
"cite_N": [
"@cite_30",
"@cite_62",
"@cite_36",
"@cite_53",
"@cite_54",
"@cite_65",
"@cite_32",
"@cite_55",
"@cite_56",
"@cite_0",
"@cite_17",
"@cite_46",
"@cite_16",
"@cite_13",
"@cite_11"
],
"mid": [
"2575842049",
"1514535095",
"2963267809",
"1895577753",
"2798959609",
"2109586012",
"2951805548",
"2963448089",
"2745461083",
"1811254738",
"2822349497",
"2625940279",
"92662927",
"2963758027",
"2607579284"
],
"abstract": [
"Attention-based neural encoder-decoder frameworks have been widely adopted for image captioning. Most methods force visual attention to be active for every generated word. However, the decoder likely requires little to no visual information from the image to predict non-visual words such as the and of. Other words that may seem visual can often be predicted reliably just from the language model e.g., sign after behind a red stop or phone following talking on a cell. In this paper, we propose a novel adaptive attention model with a visual sentinel. At each time step, our model decides whether to attend to the image (and if so, to which regions) or to the visual sentinel. The model decides whether to attend to the image and where, in order to extract meaningful information for sequential word generation. We test our method on the COCO image captioning 2015 challenge dataset and Flickr30K. Our approach sets the new state-of-the-art by a significant margin.",
"Inspired by recent work in machine translation and object detection, we introduce an attention based model that automatically learns to describe the content of images. We describe how we can train this model in a deterministic manner using standard backpropagation techniques and stochastically by maximizing a variational lower bound. We also show through visualization how the model is able to automatically learn to fix its gaze on salient objects while generating the corresponding words in the output sequence. We validate the use of attention with state-of-the-art performance on three benchmark datasets: Flickr9k, Flickr30k and MS COCO.",
"Linguistic style is an essential part of written communication, with the power to affect both clarity and attractiveness. With recent advances in vision and language, we can start to tackle the problem of generating image captions that are both visually grounded and appropriately styled. Existing approaches either require styled training captions aligned to images or generate captions with low relevance. We develop a model that learns to generate visually relevant styled captions from a large corpus of styled text without aligned images. The core idea of this model, called SemStyle, is to separate semantics and style. One key component is a novel and concise semantic term representation generated using natural language processing techniques and frame semantics. In addition, we develop a unified language model that decodes sentences with diverse word choices and syntax for different styles. Evaluations, both automatic and manual, show captions from SemStyle preserve image semantics, are descriptive, and are style shifted. More broadly, this work provides possibilities to learn richer image descriptions from the plethora of linguistic data available on the web.",
"Automatically describing the content of an image is a fundamental problem in artificial intelligence that connects computer vision and natural language processing. In this paper, we present a generative model based on a deep recurrent architecture that combines recent advances in computer vision and machine translation and that can be used to generate natural sentences describing an image. The model is trained to maximize the likelihood of the target description sentence given the training image. Experiments on several datasets show the accuracy of the model and the fluency of the language it learns solely from image descriptions. Our model is often quite accurate, which we verify both qualitatively and quantitatively. For instance, while the current state-of-the-art BLEU-1 score (the higher the better) on the Pascal dataset is 25, our approach yields 59, to be compared to human performance around 69. We also show BLEU-1 score improvements on Flickr30k, from 56 to 66, and on SBU, from 19 to 28. Lastly, on the newly released COCO dataset, we achieve a BLEU-4 of 27.7, which is the current state-of-the-art.",
"Most image captioning models focus on one-line (single image) captioning, where the correlations like relevance and diversity among group images (e.g., within the same album or event) are simply neglected, resulting in less accurate and diverse captions. Recent works mainly consider imposing the diversity during the online inference only, which neglect the correlation among visual structures in offline training. In this paper, we propose a novel group-based image captioning scheme (termed GroupCap), which jointly models the structured relevance and diversity among group images towards an optimal collaborative captioning. In particular, we first propose a visual tree parser (VP-Tree) to construct the structured semantic correlations within individual images. Then, the relevance and diversity among images are well modeled by exploiting the correlations among their tree structures. Finally, such correlations are modeled as constraints and sent into the LSTM-based captioning generator. We adopt an end-to-end formulation to train the visual tree parser, the structured relevance and diversity constraints, as well as the LSTM based captioning model jointly. To facilitate quantitative evaluation, we further release two group captioning datasets derived from the MS-COCO benchmark, serving as the first of their kind. Quantitative results show that the proposed GroupCap model outperforms the state-of-the-art and alternative approaches.",
"We develop and demonstrate automatic image description methods using a large captioned photo collection. One contribution is our technique for the automatic collection of this new dataset – performing a huge number of Flickr queries and then filtering the noisy results down to 1 million images with associated visually relevant captions. Such a collection allows us to approach the extremely challenging problem of description generation using relatively simple non-parametric methods and produces surprisingly effective results. We also develop methods incorporating many state of the art, but fairly noisy, estimates of image content to produce even more pleasing results. Finally we introduce a new objective performance measure for image captioning.",
"We present a model that generates natural language descriptions of images and their regions. Our approach leverages datasets of images and their sentence descriptions to learn about the inter-modal correspondences between language and visual data. Our alignment model is based on a novel combination of Convolutional Neural Networks over image regions, bidirectional Recurrent Neural Networks over sentences, and a structured objective that aligns the two modalities through a multimodal embedding. We then describe a Multimodal Recurrent Neural Network architecture that uses the inferred alignments to learn to generate novel descriptions of image regions. We demonstrate that our alignment model produces state of the art results in retrieval experiments on Flickr8K, Flickr30K and MSCOCO datasets. We then show that the generated descriptions significantly outperform retrieval baselines on both full images and on a new dataset of region-level annotations.",
"One property that remains lacking in image captions generated by contemporary methods is discriminability: being able to tell two images apart given the caption for one of them. We propose a way to improve this aspect of caption generation. By incorporating into the captioning training objective a loss component directly related to ability (by a machine) to disambiguate image caption matches, we obtain systems that produce much more discriminative caption, according to human evaluation. Remarkably, our approach leads to improvement in other aspects of generated captions, reflected by a battery of standard scores such as BLEU, SPICE etc. Our approach is modular and can be applied to a variety of model loss combinations commonly proposed for image captioning.",
"Top-down visual attention mechanisms have been used extensively in image captioning and visual question answering (VQA) to enable deeper image understanding through fine-grained analysis and even multiple steps of reasoning. In this work, we propose a combined bottom-up and top-down attention mechanism that enables attention to be calculated at the level of objects and other salient image regions. This is the natural basis for attention to be considered. Within our approach, the bottom-up mechanism (based on Faster R-CNN) proposes image regions, each with an associated feature vector, while the top-down mechanism determines feature weightings. Applying this approach to image captioning, our results on the MSCOCO test server establish a new state-of-the-art for the task, achieving CIDEr SPICE BLEU-4 scores of 117.9, 21.5 and 36.9, respectively. Demonstrating the broad applicability of the method, applying the same approach to VQA we obtain first place in the 2017 VQA Challenge.",
"In this paper, we present a multimodal Recurrent Neural Network (m-RNN) model for generating novel image captions. It directly models the probability distribution of generating a word given previous words and an image. Image captions are generated by sampling from this distribution. The model consists of two sub-networks: a deep recurrent neural network for sentences and a deep convolutional network for images. These two sub-networks interact with each other in a multimodal layer to form the whole m-RNN model. The effectiveness of our model is validated on four benchmark datasets (IAPR TC-12, Flickr 8K, Flickr 30K and MS COCO). Our model outperforms the state-of-the-art methods. In addition, we apply the m-RNN model to retrieval tasks for retrieving images or sentences, and achieves significant performance improvement over the state-of-the-art methods which directly optimize the ranking objective function for retrieval. The project page of this work is: www.stat.ucla.edu junhua.mao m-RNN.html .",
"Generating stylized captions for an image is an emerging topic in image captioning. Given an image as input, it requires the system to generate a caption that has a specific style (e.g., humorous, romantic, positive, and negative) while describing the image content semantically accurately. In this paper, we propose a novel stylized image captioning model that effectively takes both requirements into consideration. To this end, we first devise a new variant of LSTM, named style-factual LSTM, as the building block of our model. It uses two groups of matrices to capture the factual and stylized knowledge, respectively, and automatically learns the word-level weights of the two groups based on previous context. In addition, when we train the model to capture stylized elements, we propose an adaptive learning approach based on a reference factual model, it provides factual knowledge to the model as the model learns from stylized caption labels, and can adaptively compute how much information to supply at each time step. We evaluate our model on two stylized image captioning datasets, which contain humorous romantic captions and positive negative captions, respectively. Experiments shows that our proposed model outperforms the state-of-the-art approaches, without using extra ground truth supervision.",
"We propose a novel framework named StyleNet to address the task of generating attractive captions for images and videos with different styles. To this end, we devise a novel model component, named factored LSTM, which automatically distills the style factors in the monolingual text corpus. Then at runtime, we can explicitly control the style in the caption generation process so as to produce attractive visual captions with the desired style. Our approach achieves this goal by leveraging two sets of data: 1) factual image video-caption paired data, and 2) stylized monolingual text data (e.g., romantic and humorous sentences). We show experimentally that StyleNet outperforms existing approaches for generating visual captions with different styles, measured in both automatic and human evaluation metrics on the newly collected FlickrStyle10K image caption dataset, which contains 10K Flickr images with corresponding humorous and romantic captions.",
"This paper studies the problem of associating images with descriptive sentences by embedding them in a common latent space. We are interested in learning such embeddings from hundreds of thousands or millions of examples. Unfortunately, it is prohibitively expensive to fully annotate this many training images with ground-truth sentences. Instead, we ask whether we can learn better image-sentence embeddings by augmenting small fully annotated training sets with millions of images that have weak and noisy annotations (titles, tags, or descriptions). After investigating several state-of-the-art scalable embedding methods, we introduce a new algorithm called Stacked Auxiliary Embedding that can successfully transfer knowledge from millions of weakly annotated images to improve the accuracy of retrieval-based image description.",
"We introduce the dense captioning task, which requires a computer vision system to both localize and describe salient regions in images in natural language. The dense captioning task generalizes object detection when the descriptions consist of a single word, and Image Captioning when one predicted region covers the full image. To address the localization and description task jointly we propose a Fully Convolutional Localization Network (FCLN) architecture that processes an image with a single, efficient forward pass, requires no external regions proposals, and can be trained end-to-end with a single round of optimization. The architecture is composed of a Convolutional Network, a novel dense localization layer, and Recurrent Neural Network language model that generates the label sequences. We evaluate our network on the Visual Genome dataset, which comprises 94,000 images and 4,100,000 region-grounded captions. We observe both speed and accuracy improvements over baselines based on current state of the art approaches in both generation and retrieval settings.",
"We address personalization issues of image captioning, which have not been discussed yet in previous research. For a query image, we aim to generate a descriptive sentence, accounting for prior knowledge such as the users active vocabularies in previous documents. As applications of personalized image captioning, we tackle two post automation tasks: hashtag prediction and post generation, on our newly collected Instagram dataset, consisting of 1.1M posts from 6.3K users. We propose a novel captioning model named Context Sequence Memory Network (CSMN). Its unique updates over previous memory network models include (i) exploiting memory as a repository for multiple types of context information, (ii) appending previously generated words into memory to capture long-term information without suffering from the vanishing gradient problem, and (iii) adopting CNN memory structure to jointly represent nearby ordered memory slots for better context understanding. With quantitative evaluation and user studies via Amazon Mechanical Turk, we show the effectiveness of the three novel features of CSMN and its performance enhancement for personalized image captioning over state-of-the-art captioning models."
]
} |
1904.05732 | 2937608662 | The Kaczmarz algorithm is an iterative method for solving systems of linear equations. We introduce a modified Kaczmarz algorithm for solving systems of linear equations in a distributed environment, i.e. the equations within the system are distributed over multiple nodes within a network. The modification we introduce is designed for a network with a tree structure that allows for passage of solution estimates between the nodes in the network. We prove that the modified algorithm converges under no additional assumptions on the equations. We demonstrate that the algorithm converges to the solution, or the solution of minimal norm, when the system is consistent. We also demonstrate that in the case of an inconsistent system of equations, the modified relaxed Kaczmarz algorithm converges to a weighted least squares solution as the relaxation parameter approaches @math . | The Kaczmarz method was originally introduced in ( @cite_28 , 1937). It became popular with the introduction of Computer Tomography, under the name of ART (Algebraic Reconstruction Technique). ART added non-negativity and other constraints to the standard algorithm @cite_22 . Other variations on the Kaczmarz method allowed for relaxation parameters @cite_6 , re-ordering equations to speed up convergence @cite_10 , or considering block versions of the Kaczmarz method with relaxation matrices @math @cite_8 . Relatively recently, choosing the next equation randomly has been shown to dramatically improve the rate of convergence of the algorithm @cite_0 @cite_21 @cite_3 . Moreover, this randomized version of the Kaczmarz algorithm has been shown to comparable to the gradient descent method @cite_15 . Our version of the Kaczmarz method differs from these in that the next equation cannot be chosen randomly or otherwise, since the ordering of the equations is determined by the network topology. | {
"cite_N": [
"@cite_22",
"@cite_8",
"@cite_28",
"@cite_21",
"@cite_6",
"@cite_3",
"@cite_0",
"@cite_15",
"@cite_10"
],
"mid": [
"1990919278",
"2053342145",
"",
"1977769089",
"2144932012",
"1754711404",
"2048305372",
"2950132609",
"2075350383"
],
"abstract": [
"Abstract We give a new method for direct reconstruction of three-dimensional objects from a few electron micrographs taken at angles which need not exceed a range of 60 degrees. The method works for totally asymmetric objects, and requires little computer time or storage. It is also applicable to X-ray photography, and may greatly reduce the exposure compared to current methods of body-section radiography.",
"Abstract We present a unifying framework for a wide class of iterative methods in numerical linear algebra. In particular, the class of algorithms contains Kaczmarz's and Richardson's methods for the regularized weighted least squares problem with weighted norm. The convergence theory for this class of algorithms yields as corollaries the usual convergence conditions for Kaczmarz's and Richardson's methods. The algorithms in the class may be characterized as being group-iterative, and incorporate relaxation matrices, as opposed to a single relaxation parameter. We show that some well-known iterative methods of image reconstruction fall into the class of algorithms under consideration, and are thus covered by the convergence theory. We also describe a novel application to truly three-dimensional image reconstruction.",
"",
"The block Kaczmarz method is an iterative scheme for solving overdetermined least-squares problems. At each step, the algorithm projects the current iterate onto the solution space of a subset of the constraints. This paper describes a block Kaczmarz algorithm that uses a randomized control scheme to choose the subset at each step. This algorithm is the first block Kaczmarz method with an (expected) linear rate of convergence that can be expressed in terms of the geometric properties of the matrix and its submatrices. The analysis reveals that the algorithm is most effective when it is given a good row paving of the matrix, a partition of the rows into well-conditioned blocks. The operator theory literature provides detailed information about the existence and construction of good row pavings. Together, these results yield an efficient block Kaczmarz scheme that applies to many overdetermined least-squares problem.",
"The iterative method for solving system of linear equations, due to Kaczmarz [2], is investigated. It is shown that the method works well for both singular and non-singular systems and it determines the affine space formed by the solutions if they exist. The method also provides an iterative procedure for computing a generalized inverse of a matrix.",
"Abstract The Kaczmarz method is an iterative method for solving overcomplete linear systems of equations A x = b . The randomized version of the Kaczmarz method put forth by Strohmer and Vershynin iteratively projects onto a randomly chosen solution space given by a single row of the matrix A and converges linearly 1 in expectation to the solution of a consistent system. In this paper we analyze two block versions of the method each with a randomized projection, designed to converge in expectation to the least squares solution, often faster than the standard variants. Our approach utilizes both a row and column-paving of the matrix A to guarantee linear convergence when the matrix has consistent row norms (called nearly standardized), and a single column-paving when the row norms are unbounded. The proposed methods are an extension of the block Kaczmarz method analyzed by Needell and Tropp and the Randomized Extended Kaczmarz method of Zouzias and Freris. The contribution is thus two-fold; unlike the standard Kaczmarz method, our results demonstrate convergence to the least squares solution of inconsistent systems (both methods in the nearly standardized case and the second method in other cases). By using appropriate blocks of the matrix this convergence can be significantly accelerated, as is demonstrated by numerical experiments.",
"The Kaczmarz method for solving linear systems of equations is an iterative algorithm that has found many applications ranging from computer tomography to digital signal processing. Despite the popularity of this method, useful theoretical estimates for its rate of convergence are still scarce. We introduce a randomized version of the Kaczmarz method for consistent, overdetermined linear systems and we prove that it converges with expected exponential rate. Furthermore, this is the first solver whose rate does not depend on the number of equations in the system. The solver does not even need to know the whole system but only a small random part of it. It thus outperforms all previously known methods on general extremely overdetermined systems. Even for moderately overdetermined systems, numerical simulations as well as theoretical analysis reveal that our algorithm can converge faster than the celebrated conjugate gradient algorithm. Furthermore, our theory and numerical simulations confirm a prediction of in the context of reconstructing bandlimited functions from nonuniform sampling.",
"We obtain an improved finite-sample guarantee on the linear convergence of stochastic gradient descent for smooth and strongly convex objectives, improving from a quadratic dependence on the conditioning @math (where @math is a bound on the smoothness and @math on the strong convexity) to a linear dependence on @math . Furthermore, we show how reweighting the sampling distribution (i.e. importance sampling) is necessary in order to further improve convergence, and obtain a linear dependence in the average smoothness, dominating previous results. We also discuss importance sampling for SGD more broadly and show how it can improve convergence also in other scenarios. Our results are based on a connection we make between SGD and the randomized Kaczmarz algorithm, which allows us to transfer ideas between the separate bodies of literature studying each of the two methods. In particular, we recast the randomized Kaczmarz algorithm as an instance of SGD, and apply our results to prove its exponential convergence, but to the solution of a weighted least squares problem rather than the original least squares problem. We then present a modified Kaczmarz algorithm with partially biased sampling which does converge to the original least squares solution with the same exponential convergence rate.",
"Spring balance apparatus including a weighing lever supported on a knife edge balance point with one side of the lever acted on by a mass to be weighed against a weighing spring coupled with the other side of the weighing lever, and including an arrangement for the compensation of a measuring error due to temperature change by modifying the ratio of transmission of forces applied to the weighing lever by means of a bimet allic element in which the inventive improvement comprises a knife edge member transferring the forces of the weighing spring to the other side of the weighing lever which knife-edge member has connected rigidly to it an actuating lever coupled to the bimet allic element to alter responsive to temperature changes the orientation of the knife edge member relative to the other side of the weighing lever and thereby modify the ratio of transmission of forces to compensate for measuring error that would otherwise occur due to temperature change."
]
} |
1904.05732 | 2937608662 | The Kaczmarz algorithm is an iterative method for solving systems of linear equations. We introduce a modified Kaczmarz algorithm for solving systems of linear equations in a distributed environment, i.e. the equations within the system are distributed over multiple nodes within a network. The modification we introduce is designed for a network with a tree structure that allows for passage of solution estimates between the nodes in the network. We prove that the modified algorithm converges under no additional assumptions on the equations. We demonstrate that the algorithm converges to the solution, or the solution of minimal norm, when the system is consistent. We also demonstrate that in the case of an inconsistent system of equations, the modified relaxed Kaczmarz algorithm converges to a weighted least squares solution as the relaxation parameter approaches @math . | A distributed version of the Kaczmarz algorithm was introduced in @cite_20 . The main ideas presented there are very similar to ours: updated estimates are obtained from prior estimates using the Kaczmarz update with the equations that are available at the node, and distributed estimates are averaged together at a single node (which the authors refer to as a fusion center, for us it is the root of the tree). In @cite_20 , the convergence analysis is limited to the case of consistent systems of equations, and inconsistent systems are handled by Tikhonov regularization @cite_17 @cite_26 rather than by varying the relaxation parameter. | {
"cite_N": [
"@cite_26",
"@cite_20",
"@cite_17"
],
"mid": [
"",
"1496894444",
"2083268403"
],
"abstract": [
"",
"Many real-world wireless sensor network applications such as environmental monitoring, structural health monitoring, and smart grid can be formulated as a least-squares problem. In distributed Cyber-Physical System (CPS), each sensor node observes partial phenomena due to spatial and temporal restriction and is able to form only partial rows of least-squares. Traditionally, these partial measurements were gathered at a centralized location. However, with the increase in sensors and their measurements, aggregation is becoming challenging and infeasible. In this paper, we propose distributed randomized kaczmarz that performs in-network computation to solve least-squares over the network by avoiding costly communication. As a case study, we present a volcano monitoring application on a distributed CORE emulator and use real data from Mt. St. Helens to evaluate our proposed method.",
"A new iterative method is proposed for finding the optimal Bayesian estimate of an unknown image from its projection data (experimentally obtained integrals of its grayness over thin strips). Convergence of the method is proved and its performance is illustrated. The method compares favorably with previously proposed procedures."
]
} |
1904.05606 | 2938532952 | This paper deals with multi-lingual dialogue act (DA) recognition. The proposed approaches are based on deep neural networks and use word2vec embeddings for word representation. Two multi-lingual models are proposed for this task. The first approach uses one general model trained on the embeddings from all available languages. The second method trains the model on a single pivot language and a linear transformation method is used to project other languages onto the pivot language. The popular convolutional neural network and LSTM architectures with different set-ups are used as classifiers. To the best of our knowledge this is the first attempt at multi-lingual DA recognition using neural networks. The multi-lingual models are validated experimentally on two languages from the Verbmobil corpus. | Several lexical models are proposed including Bayesian approaches such as n-gram language models @cite_23 @cite_4 @cite_3 and also non-Bayesian methods -- semantic classification trees @cite_12 , transformation-based learning (TBL) @cite_0 or memory-based learning @cite_13 . Syntactic features that are created using a full parse tree are considered for instance in @cite_14 . Prosodic information is often used to provide additional clues to recognize DAs as presented for instance in @cite_27 . Dialogue history (the sequence of DAs) can been used to predict the most probable next dialogue act and it can be modeled for example by hidden Markov models @cite_4 or Bayesian networks @cite_28 . | {
"cite_N": [
"@cite_14",
"@cite_4",
"@cite_28",
"@cite_3",
"@cite_0",
"@cite_27",
"@cite_23",
"@cite_13",
"@cite_12"
],
"mid": [
"2079954222",
"2128970689",
"2030221669",
"2146785422",
"2170927607",
"2112078429",
"155915536",
"99009221",
"2180081471"
],
"abstract": [
"This work studies the usefulness of syntactic information in the context of automatic dialogue act recognition in Czech. Several pieces of evidence are presented in this work that support our claim that syntax might bring valuable information for dialogue act recognition. In particular, a parallel is drawn with the related domain of automatic punctuation generation and a set of syntactic features derived from a deep parse tree is further proposed and successfully used in a Czech dialogue act recognition system based on conditional random fields. We finally discuss the possible reasons why so few works have exploited this type of information before and propose future research directions to further progress in this area.",
"We describe a statistical approach for modeling dialogue acts in conversational speech, i.e., speech-act-like units such as STATEMENT, QUESTION, BACKCHANNEL, AGREEMENT, DISAGREEMENT, and APOLOGY. Our model detects and predicts dialogue acts based on lexical, collocational, and prosodic cues, as well as on the discourse coherence of the dialogue act sequence. The dialogue model is based on treating the discourse structure of a conversation as a hidden Markov model and the individual dialogue acts as observations emanating from the model states. Constraints on the likely sequence of dialogue acts are modeled via a dialogue act n-gram. The statistical dialogue grammar is combined with word n-grams, decision trees, and neural networks modeling the idiosyncratic lexical and prosodic manifestations of each dialogue act. We develop a probabilistic integration of speech recognition with dialogue modeling, to improve both speech recognition and dialogue act classification accuracy. Models are trained and evaluated using a large hand-labeled database of 1,155 conversations from the Switchboard corpus of spontaneous human-to-human telephone speech. We achieved good dialogue act labeling accuracy (65 based on errorful, automatically recognized words and prosody, and 71 based on word transcripts, compared to a chance baseline accuracy of 35 and human accuracy of 84 ) and a small reduction in word recognition error.",
"",
"We explore the two related tasks of dialog act (DA) segmentation and DA classification for speech from the ICSI Meeting Corpus. We employ simple lexical and prosodic knowledge sources, and compare results for human-transcribed versus automatically recognized words. Since there is little previous work on DA segmentation and classification in the meeting domain, our study provides baseline performance rates for both tasks. We introduce a range of metrics for use in evaluation, each of which measures different aspects of interest. Results show that both tasks are difficult, particularly for a fully automatic system. We find that a very simple prosodic model aids performance over lexical information alone, especially for segmentation. Both tasks, but particularly word-based segmentation, are degraded by word recognition errors. Finally, while classification results for meeting data show some similarities to previous results for telephone conversations, findings also suggest a potential difference with respect to the effect of modeling DA context.",
"For the task of recognizing dialogue acts, we are applying the Transformation-Based Learning (TBL) machine learning algorithm. To circumvent a sparse data problem, we extract values of well-motivated features of utterances, such as speaker direction, punctuation marks, and a new feature, called dialogue act cues, which we find to be more effective than cue phrases and word n-grams in practice. We present strategies for constructing a set of dialogue act cues automatically by minimizing the entropy of the distribution of dialogue acts in a training corpus, filtering out irrelevant dialogue act cues, and clustering semantically-related words. In addition, to address limitations of TBL, we introduce a Monte Carlo strategy for training efficiently and a committee method for computing confidence measures. These ideas are combined in our working implementation, which labels held-out data as accurately as any other reported system for the dialogue act tagging task.",
"Identifying whether an utterance is a statement, question, greeting, and so forth is integral to effective automatic understanding of natural dialog. Little is known, however, about how such dialog acts (DAs) can be automatically classified in truly natural conversation. This study asks whether current approaches, which use mainly word information, could be improved by adding prosodic information.The study is based on more than 1000 conversations from the Switchboard corpus. DAs were hand-annotated, and prosodic features (duration, pause, F0, energy, and speaking rate) were automatically extracted for each DA. In training, decision trees based on these features were inferred; trees were then applied to unseen test data to evaluate performance. Performance was evaluated for prosody models alone, and after combining the prosody models with word information—either from true words or from the output of an automatic speech recognizer.For an overall classification task, as well as three subtasks, prosody made s...",
"",
"Ligand-binding, dimerized polypeptide fusions comprising two polypeptide chains are disclosed. Each of the polypeptide chains comprises a ligand-binding domain of either a growth factor receptor or a hormone receptor joined to an immunoglobulin constant region polypeptide. Dimerization of the polypeptide fusion results in enhancement of biological activity. The immunoglobulin constant region polypeptide may comprise an immunoglobulin heavy chain constant region domain and hinge region.",
"This paper presents automatic methods for the classification of dialog acts. In the verbmobil application (speech-to-speech translation of face-to-face dialogs) maximally 50 of the utterances are analyzed in depth and for the rest, shallow processing takes place. The dialog component keeps track of the dialog with this shallow processing. For the classification of utterances without in depth processing two methods are presented: Semantic Classification Trees and Polygrams. For both methods the classification algorithm is trained automatically from a corpus of labeled data. The novel idea with respect to SCTs is the use of dialog state dependent CTs and with respect to Polygrams it is the use of competing language models for the classification of dialog acts."
]
} |
1904.05606 | 2938532952 | This paper deals with multi-lingual dialogue act (DA) recognition. The proposed approaches are based on deep neural networks and use word2vec embeddings for word representation. Two multi-lingual models are proposed for this task. The first approach uses one general model trained on the embeddings from all available languages. The second method trains the model on a single pivot language and a linear transformation method is used to project other languages onto the pivot language. The popular convolutional neural network and LSTM architectures with different set-ups are used as classifiers. To the best of our knowledge this is the first attempt at multi-lingual DA recognition using neural networks. The multi-lingual models are validated experimentally on two languages from the Verbmobil corpus. | Nowadays, DA recognition is often realized with simple word features and sophisticated neural models including popular CNNs and an LSTM. The features are represented by word embeddings provided by word2vec @cite_22 or glove @cite_26 . Both word2vec and glove are used in @cite_2 where a CNN (or an LSTM) is used to create a vector representing a DA. This vector is then fed into a feed forward network utilized for classification. The experiments on several DA corpora have shown that the CNN performs slightly better than the LSTM in this case. Another approach using only raw word forms as an input is presented in @cite_11 . The LSTM with word2vec embeddings is used to model the DA structure as well as for DA prediction with very good results. A similar LSTM model with word2vec embeddings has been also proposed for DA recognition in @cite_8 . The authors experimented with different embedding sizes and network hyper-parameters. @cite_15 propose new features called "probabilistic word embeddings" which are based on word distribution across DA-set. The experiments show that these features perform slightly better than word2vec. | {
"cite_N": [
"@cite_26",
"@cite_22",
"@cite_8",
"@cite_2",
"@cite_15",
"@cite_11"
],
"mid": [
"2250539671",
"1614298861",
"2573626026",
"2964331270",
"2883617287",
"2740155805"
],
"abstract": [
"Recent methods for learning vector space representations of words have succeeded in capturing fine-grained semantic and syntactic regularities using vector arithmetic, but the origin of these regularities has remained opaque. We analyze and make explicit the model properties needed for such regularities to emerge in word vectors. The result is a new global logbilinear regression model that combines the advantages of the two major model families in the literature: global matrix factorization and local context window methods. Our model efficiently leverages statistical information by training only on the nonzero elements in a word-word cooccurrence matrix, rather than on the entire sparse matrix or on individual context windows in a large corpus. The model produces a vector space with meaningful substructure, as evidenced by its performance of 75 on a recent word analogy task. It also outperforms related models on similarity tasks and named entity recognition.",
"",
"This paper applies a deep long-short term memory (LSTM) structure to classify dialogue acts in open-domain conversations.",
"",
"The identification of Dialogue Act’s (DA) is an important aspect in determining the meaning of an utterance for many applications that require natural language understanding, and recent work using recurrent neural networks (RNN) has shown promising results when applied to the DA classification problem. This work presents a novel probabilistic method of utterance representation and describes a RNN sentence model for out-of-context DA Classification. The utterance representations are generated from keywords selected for their frequency association with certain DA’s. The proposed probabilistic representations are applied to the Switchboard DA corpus and performance is compared with pre-trained word embeddings using the same baseline RNN model. The results indicate that the probabilistic method achieves 75.48 overall accuracy and an improvement over the word embedding representations of 1.8 . This demonstrates the potential utility of using statistical utterance representations, that are able to capture word-DA relationships, for the purpose of DA classification.",
"Abstract Dialogue act recognition is an important component of a large number of natural language processing pipelines. Many research works have been carried out in this area, but relatively few investigate deep neural networks and word embeddings. This is surprising, given that both of these techniques have proven exceptionally good in most other language-related domains. We propose in this work a new deep neural network that explores recurrent models to capture word sequences within sentences, and further study the impact of pretrained word embeddings. We validate this model on three languages: English, French and Czech. The performance of the proposed approach is consistent across these languages and it is comparable to the state-of-the-art results in English. More importantly, we confirm that deep neural networks indeed outperform a Maximum Entropy classifier, which was expected. However, and this is more surprising, we also found that standard word2vec embeddings do not seem to bring valuable information for this task and the proposed model, whatever the size of the training corpus is. We thus further analyse the resulting embeddings and conclude that a possible explanation may be related to the mismatch between the type of lexical-semantic information captured by the word2vec embeddings, and the kind of relations between words that is the most useful for the dialogue act recognition task."
]
} |
1904.05606 | 2938532952 | This paper deals with multi-lingual dialogue act (DA) recognition. The proposed approaches are based on deep neural networks and use word2vec embeddings for word representation. Two multi-lingual models are proposed for this task. The first approach uses one general model trained on the embeddings from all available languages. The second method trains the model on a single pivot language and a linear transformation method is used to project other languages onto the pivot language. The popular convolutional neural network and LSTM architectures with different set-ups are used as classifiers. To the best of our knowledge this is the first attempt at multi-lingual DA recognition using neural networks. The multi-lingual models are validated experimentally on two languages from the Verbmobil corpus. | The above mentioned approaches are mainly evaluated on English language using Switchboard (SwDA) @cite_24 or Meeting Recorder Dialog Act (MRDA) @cite_5 corpora. Some methods are tested on Spanish (see DIHANA @cite_20 corpus), Czech @cite_25 , French @cite_16 or on German (see Verbmobil @cite_7 corpus) languages. | {
"cite_N": [
"@cite_7",
"@cite_24",
"@cite_5",
"@cite_16",
"@cite_25",
"@cite_20"
],
"mid": [
"",
"",
"1503312748",
"2251052362",
"60398317",
"70435779"
],
"abstract": [
"",
"",
"Abstract : We describe a new corpus of over 180,000 hand- annotated dialog act tags and accompanying adjacency pair annotations for roughly 72 hours of speech from 75 naturally-occurring meetings. We provide a brief summary of the annotation system and labeling procedure, inter-annotator reliability statistics, overall distributional statistics, a description of auxiliary files distributed with the corpus, and information on how to obtain the data.",
"We describe the acquisition of a dialog corpus for French based on multi-task human-machine interactions in a serious game setting. We present a tool for data collection that is configurable for multiple games; describe the data collected using this tool and the annotation schema used to annotate it; and report on the results obtained when training a classifier on the annotated data to associate each player turn with a dialog move usable by a rule based dialog manager. The collected data consists of approximately 1250 dialogs, 10454 utterances and 168509 words and will be made freely available to academic and nonprofit research.",
"This paper deals with automatic dialog acts (DAs) recognition in Czech. The dialog acts are sentence-level labels that represent different states of a dialogue, depending on the application. Our work focuses on two applications: a multimodal reservation system and an animated talking head for hearing-impaired people. In that context, we consider the following DAs: statements, orders, yes no questions and other questions. We propose to use both lexical and prosodic information for DAs recognition. The main goal of this paper is to compare different methods to combine the results of both classifiers. On a Czech corpus simulating a reservation of train tickets, the lexical information only gives about 92 of classification accuracy, while prosody gives only about 45 of accuracy. When both classifiers are combined with a multilayer perceptron, the lowest (lexical) word error rate further decreases by 26 . We show that this improvement is close to the optimal one, given the correlation of the lexical and prosodic features. The other combination schemes do not outperform the lexical-only results.",
"In the framework of the DIHANA project, we present the acquisition process of a spontaneous speech dialogue corpus in Spanish. The selected application consists of information retrieval by telephone for nationwide trains. A total of 900 dialogues from 225 users were acquired using the Wizard of Oz technique. In this work, we present the design and planning of the dialogue scenes and the wizard strategy used for the acquisition of the corpus. Then, we also present the acquisition tools and a description of the acquisition process."
]
} |
1904.05606 | 2938532952 | This paper deals with multi-lingual dialogue act (DA) recognition. The proposed approaches are based on deep neural networks and use word2vec embeddings for word representation. Two multi-lingual models are proposed for this task. The first approach uses one general model trained on the embeddings from all available languages. The second method trains the model on a single pivot language and a linear transformation method is used to project other languages onto the pivot language. The popular convolutional neural network and LSTM architectures with different set-ups are used as classifiers. To the best of our knowledge this is the first attempt at multi-lingual DA recognition using neural networks. The multi-lingual models are validated experimentally on two languages from the Verbmobil corpus. | We further implemented another baseline that uses a maximum entropy (ME) classifier with simple bag of words (BoW) features. Then, we show the results of our previous LSTM system @cite_11 , which uses only simple word level features and word tokens from the previous dialogue act. The results of our approaches are presented in the last two lines of this table. | {
"cite_N": [
"@cite_11"
],
"mid": [
"2740155805"
],
"abstract": [
"Abstract Dialogue act recognition is an important component of a large number of natural language processing pipelines. Many research works have been carried out in this area, but relatively few investigate deep neural networks and word embeddings. This is surprising, given that both of these techniques have proven exceptionally good in most other language-related domains. We propose in this work a new deep neural network that explores recurrent models to capture word sequences within sentences, and further study the impact of pretrained word embeddings. We validate this model on three languages: English, French and Czech. The performance of the proposed approach is consistent across these languages and it is comparable to the state-of-the-art results in English. More importantly, we confirm that deep neural networks indeed outperform a Maximum Entropy classifier, which was expected. However, and this is more surprising, we also found that standard word2vec embeddings do not seem to bring valuable information for this task and the proposed model, whatever the size of the training corpus is. We thus further analyse the resulting embeddings and conclude that a possible explanation may be related to the mismatch between the type of lexical-semantic information captured by the word2vec embeddings, and the kind of relations between words that is the most useful for the dialogue act recognition task."
]
} |
1904.05547 | 2940165378 | 3D human pose estimation from a monocular image or 2D joints is an ill-posed problem because of depth ambiguity and occluded joints. We argue that 3D human pose estimation from a monocular input is an inverse problem where multiple feasible solutions can exist. In this paper, we propose a novel approach to generate multiple feasible hypotheses of the 3D pose from 2D joints.In contrast to existing deep learning approaches which minimize a mean square error based on an unimodal Gaussian distribution, our method is able to generate multiple feasible hypotheses of 3D pose based on a multimodal mixture density networks. Our experiments show that the 3D poses estimated by our approach from an input of 2D joints are consistent in 2D reprojections, which supports our argument that multiple solutions exist for the 2D-to-3D inverse problem. Furthermore, we show state-of-the-art performance on the Human3.6M dataset in both best hypothesis and multi-view settings, and we demonstrate the generalization capacity of our model by testing on the MPII and MPI-INF-3DHP datasets. Our code is available at the project website. | The second category @cite_13 @cite_23 @cite_21 @cite_34 @cite_24 @cite_25 @cite_31 @cite_32 decouples 3D pose estimation into the well-studied 2D joint detection @cite_27 @cite_15 and 3D pose estimation from the detected 2D joints. Akhter al @cite_13 propose a multi-stage approach to estimate the 3D pose from 2D joints using an over-complete dictionary of poses. Bogo al @cite_19 estimate 3D pose by first fitting a statistical body shape model to the 2D joints, and then minimizing the error between the reprojected 3D model and detected 2D joints. Chen @cite_22 and Yasin @cite_31 regard 3D pose estimation as a matching between the estimated 2D pose and the 3D pose from a large pose library. Martinez al @cite_34 design a simple fully connected residual network to regress 3D pose from 2D joint detections. The decoupled approach can make use of both indoor and in-the-wild images to train the 2D pose estimators. More importantly, this approach is domain invariant since the input of the second stage is the 2D joints. However, estimating 3D pose from 2D joints is more challenging because 2D pose data contains less information than images, thus there are more ambiguities. | {
"cite_N": [
"@cite_22",
"@cite_15",
"@cite_21",
"@cite_32",
"@cite_24",
"@cite_19",
"@cite_27",
"@cite_23",
"@cite_31",
"@cite_34",
"@cite_13",
"@cite_25"
],
"mid": [
"2962729993",
"",
"",
"",
"2769237672",
"2483862638",
"2307770531",
"2963688992",
"2963013806",
"2612706635",
"1943191679",
""
],
"abstract": [
"Human 3D pose estimation from a single image is a challenging task with numerous applications. Convolutional Neural Networks (CNNs) have recently achieved superior performance on the task of 2D pose estimation from a single image, by training on images with 2D annotations collected by crowd sourcing. This suggests that similar success could be achieved for direct estimation of 3D poses. However, 3D poses are much harder to annotate, and the lack of suitable annotated training images hinders attempts towards end-to-end solutions. To address this issue, we opt to automatically synthesize training images with ground truth pose annotations. Our work is a systematic study along this road. We find that pose space coverage and texture diversity are the key ingredients for the effectiveness of synthetic training data. We present a fully automatic, scalable approach that samples the human pose space for guiding the synthesis procedure and extracts clothing textures from real images. Furthermore, we explore domain adaptation for bridging the gap between our synthetic training images and real testing photos. We demonstrate that CNNs trained with our synthetic images out-perform those trained with real photos on 3D pose estimation tasks.",
"",
"",
"",
"In this work, we address the problem of 3D human pose estimation from a sequence of 2D human poses. Although the recent success of deep networks has led many state-of-the-art methods for 3D pose estimation to train deep networks end-to-end to predict from images directly, the top-performing approaches have shown the effectiveness of dividing the task of 3D pose estimation into two steps: using a state-of-the-art 2D pose estimator to estimate the 2D pose from images and then mapping them into 3D space. They also showed that a low-dimensional representation like 2D locations of a set of joints can be discriminative enough to estimate 3D pose with high accuracy. However, estimation of 3D pose for individual frames leads to temporally incoherent estimates due to independent error in each frame causing jitter. Therefore, in this work we utilize the temporal information across a sequence of 2D joint locations to estimate a sequence of 3D poses. We designed a sequence-to-sequence network composed of layer-normalized LSTM units with shortcut connections connecting the input to the output on the decoder side and imposed temporal smoothness constraint during training. We found that the knowledge of temporal consistency improves the best reported result on Human3.6M dataset by approximately (12.2 ) and helps our network to recover temporally consistent 3D poses over a sequence of images even when the 2D pose detector fails.",
"We describe the first method to automatically estimate the 3D pose of the human body as well as its 3D shape from a single unconstrained image. We estimate a full 3D mesh and show that 2D joints alone carry a surprising amount of information about body shape. The problem is challenging because of the complexity of the human body, articulation, occlusion, clothing, lighting, and the inherent ambiguity in inferring 3D from 2D. To solve this, we first use a recently published CNN-based method, DeepCut, to predict (bottom-up) the 2D body joint locations. We then fit (top-down) a recently published statistical body shape model, called SMPL, to the 2D joints. We do so by minimizing an objective function that penalizes the error between the projected 3D model joints and detected 2D joints. Because SMPL captures correlations in human shape across the population, we are able to robustly fit it to very little data. We further leverage the 3D model to prevent solutions that cause interpenetration. We evaluate our method, SMPLify, on the Leeds Sports, HumanEva, and Human3.6M datasets, showing superior pose accuracy with respect to the state of the art.",
"This work introduces a novel convolutional network architecture for the task of human pose estimation. Features are processed across all scales and consolidated to best capture the various spatial relationships associated with the body. We show how repeated bottom-up, top-down processing used in conjunction with intermediate supervision is critical to improving the performance of the network. We refer to the architecture as a “stacked hourglass” network based on the successive steps of pooling and upsampling that are done to produce a final set of predictions. State-of-the-art results are achieved on the FLIC and MPII benchmarks outcompeting all recent methods.",
"This paper addresses the challenge of 3D full-body human pose estimation from a monocular image sequence. Here, two cases are considered: (i) the image locations of the human joints are provided and (ii) the image locations of joints are unknown. In the former case, a novel approach is introduced that integrates a sparsity-driven 3D geometric prior and temporal smoothness. In the latter case, the former case is extended by treating the image locations of the joints as latent variables to take into account considerable uncertainties in 2D joint locations. A deep fully convolutional network is trained to predict the uncertainty maps of the 2D joint locations. The 3D pose estimates are realized via an Expectation-Maximization algorithm over the entire sequence, where it is shown that the 2D joint location uncertainties can be conveniently marginalized out during inference. Empirical evaluation on the Human3.6M dataset shows that the proposed approaches achieve greater 3D pose estimation accuracy over state-of-the-art baselines. Further, the proposed approach outperforms a publicly available 2D pose estimation baseline on the challenging PennAction dataset.",
"One major challenge for 3D pose estimation from a single RGB image is the acquisition of sufficient training data. In particular, collecting large amounts of training data that contain unconstrained images and are annotated with accurate 3D poses is infeasible. We therefore propose to use two independent training sources. The first source consists of images with annotated 2D poses and the second source consists of accurate 3D motion capture data. To integrate both sources, we propose a dual-source approach that combines 2D pose estimation with efficient and robust 3D pose retrieval. In our experiments, we show that our approach achieves state-of-the-art results and is even competitive when the skeleton structure of the two sources differ substantially.",
"Following the success of deep convolutional networks, state-of-the-art methods for 3d human pose estimation have focused on deep end-to-end systems that predict 3d joint locations given raw image pixels. Despite their excellent performance, it is often not easy to understand whether their remaining error stems from a limited 2d pose (visual) understanding, or from a failure to map 2d poses into 3- dimensional positions.,,With the goal of understanding these sources of error, we set out to build a system that given 2d joint locations predicts 3d positions. Much to our surprise, we have found that, with current technology, \"lifting\" ground truth 2d joint locations to 3d space is a task that can be solved with a remarkably low error rate: a relatively simple deep feedforward network outperforms the best reported result by about 30 on Human3.6M, the largest publicly available 3d pose estimation benchmark. Furthermore, training our system on the output of an off-the-shelf state-of-the-art 2d detector (i.e., using images as input) yields state of the art results – this includes an array of systems that have been trained end-to-end specifically for this task. Our results indicate that a large portion of the error of modern deep 3d pose estimation systems stems from their visual analysis, and suggests directions to further advance the state of the art in 3d human pose estimation.",
"Estimating 3D human pose from 2D joint locations is central to the analysis of people in images and video. To address the fact that the problem is inherently ill posed, many methods impose a prior over human poses. Unfortunately these priors admit invalid poses because they do not model how joint-limits vary with pose. Here we make two key contributions. First, we collect a motion capture dataset that explores a wide range of human poses. From this we learn a pose-dependent model of joint limits that forms our prior. Both dataset and prior are available for research purposes. Second, we define a general parametrization of body pose and a new, multi-stage, method to estimate 3D pose from 2D joint locations using an over-complete dictionary of poses. Our method shows good generalization while avoiding impossible poses. We quantitatively compare our method with recent work and show state-of-the-art results on 2D to 3D pose estimation using the CMU mocap dataset. We also show superior results using manual annotations on real images and automatic detections on the Leeds sports pose dataset.",
""
]
} |
1904.05488 | 2939745571 | Current deep neural networks suffer from two problems; first, they are hard to interpret, and second, they suffer from overfitting. There have been many attempts to define interpretability in neural networks, but they typically lack causality or generality. A myriad of regularization techniques have been developed to prevent overfitting, and this has driven deep learning to become the hot topic it is today; however, while most regularization techniques are justified empirically and even intuitively, there is not much underlying theory. This paper argues that to extract the features used in neural networks to make decisions, it's important to look at the paths between clusters existing in the hidden spaces of neural networks. These features are of particular interest because they reflect the true decision making process of the neural network. This analysis is then furthered to present an ensemble algorithm for arbitrary neural networks which has guarantees for test accuracy. Finally, a discussion detailing the aforementioned guarantees is introduced and the implications to neural networks, including an intuitive explanation for all current regularization methods, are presented. The ensemble algorithm has generated state-of-the-art results for Wide-ResNet on CIFAR-10 and has improved test accuracy for all models it has been applied to. | Much effort has been put into interpreting deep neural networks in the past. A few meta studies summarize the effort of the community as a whole rather well. There are three general types of methods for deep neural networks; namely, by discovering "ways to reduce the complexity of all these operations," "understand[ing] the role and structure of the data flowing through these bottlenecks," and "creat[ing] networks that are designed to be easier to explain" @cite_11 . The difference between the algorithm presented here and the aforementioned attempts at interpretability is that this derives its logic entirely from the network--in particular, paths, by definition, represent the features used in classification. | {
"cite_N": [
"@cite_11"
],
"mid": [
"2963847595"
],
"abstract": [
"There has recently been a surge of work in explanatory artificial intelligence (XAI). This research area tackles the important problem that complex machines and algorithms often cannot provide insights into their behavior and thought processes. XAI allows users and parts of the internal system to be more transparent, providing explanations of their decisions in some level of detail. These explanations are important to ensure algorithmic fairness, identify potential bias problems in the training data, and to ensure that the algorithms perform as expected. However, explanations produced by these systems is neither standardized nor systematically assessed. In an effort to create best practices and identify open challenges, we describe foundational concepts of explainability and show how they can be used to classify existing literature. We discuss why current approaches to explanatory methods especially for deep neural networks are insufficient. Finally, based on our survey, we conclude with suggested future research directions for explanatory artificial intelligence."
]
} |
1904.05488 | 2939745571 | Current deep neural networks suffer from two problems; first, they are hard to interpret, and second, they suffer from overfitting. There have been many attempts to define interpretability in neural networks, but they typically lack causality or generality. A myriad of regularization techniques have been developed to prevent overfitting, and this has driven deep learning to become the hot topic it is today; however, while most regularization techniques are justified empirically and even intuitively, there is not much underlying theory. This paper argues that to extract the features used in neural networks to make decisions, it's important to look at the paths between clusters existing in the hidden spaces of neural networks. These features are of particular interest because they reflect the true decision making process of the neural network. This analysis is then furthered to present an ensemble algorithm for arbitrary neural networks which has guarantees for test accuracy. Finally, a discussion detailing the aforementioned guarantees is introduced and the implications to neural networks, including an intuitive explanation for all current regularization methods, are presented. The ensemble algorithm has generated state-of-the-art results for Wide-ResNet on CIFAR-10 and has improved test accuracy for all models it has been applied to. | In addition, clustering has been applied to the hidden spaces of neural networks in prior work, also in the context of interpretability @cite_18 . However, cluster path analysis in deep neural networks could not be found. | {
"cite_N": [
"@cite_18"
],
"mid": [
"2963041618"
],
"abstract": [
"Model interpretability is a requirement in many applications in which crucial decisions are made by users relying on a model's outputs. The recent movement for “algorithmic fairness” also stipulates explainability, and therefore interpretability of learning models. And yet the most successful contemporary Machine Learning approaches, the Deep Neural Networks, produce models that are highly non-interpretable. We attempt to address this challenge by proposing a technique called CNN-INTE to interpret deep Convolutional Neural Networks (CNN) via meta-learning. In this work, we interpret a specific hidden layer of the deep CNN model on the MNIST image dataset. We use a clustering algorithm in a two-level structure to find the meta-level training data and Random Forest as base learning algorithms to generate the meta-level test data. The interpretation results are displayed visually via diagrams, which clearly indicates how a specific test instance is classified. Our method achieves global interpretation for all the test instances on the hidden layers without sacrificing the accuracy obtained by the original deep CNN model. This means our model is faithful to the original deep CNN model, which leads to reliable interpretations."
]
} |
1904.05562 | 2938318677 | Recently, 3D face reconstruction and face alignment tasks are gradually combined into one task: 3D dense face alignment. Its goal is to reconstruct the 3D geometric structure of face with pose information. In this paper, we propose a graph convolution network to regress 3D face coordinates. Our method directly performs feature learning on the 3D face mesh, where the geometric structure and details are well preserved. Extensive experiments show that our approach gains superior performance over state-of-the-art methods on several challenging datasets. | An early representative work is 3DDFA @cite_30 . It uses a cascaded CNN to fit the 3DMM @cite_15 parameters and the projection matrix from a single 2D face image. It shows promising alignment result across large poses, and defines the new problem called 3D dense face alignment. @cite_28 utilizes multi-constraints such as face contours and SIFT feature points to estimate the 3D face shape. It provides a very dense 3D alignment. Although these 3DMM based methods achieve good performance, their accuracy is up to the capacity of the 3DMM. PRN @cite_21 uses a simple encoder-decoder architecture to regress the 3D face coordinates, which are stored in uv position map. Although it achieves state-of-the-art performance, there are many stripes on the reconstructed 3D face. This is attributed to the geometry distortion caused by uv mapping. | {
"cite_N": [
"@cite_30",
"@cite_15",
"@cite_28",
"@cite_21"
],
"mid": [
"2964014798",
"2237250383",
"2963253045",
"2963342110"
],
"abstract": [
"Face alignment, which fits a face model to an image and extracts the semantic meanings of facial pixels, has been an important topic in CV community. However, most algorithms are designed for faces in small to medium poses (below 45), lacking the ability to align faces in large poses up to 90. The challenges are three-fold: Firstly, the commonly used landmark-based face model assumes that all the landmarks are visible and is therefore not suitable for profile views. Secondly, the face appearance varies more dramatically across large poses, ranging from frontal view to profile view. Thirdly, labelling landmarks in large poses is extremely challenging since the invisible landmarks have to be guessed. In this paper, we propose a solution to the three problems in an new alignment framework, called 3D Dense Face Alignment (3DDFA), in which a dense 3D face model is fitted to the image via convolutional neutral network (CNN). We also propose a method to synthesize large-scale training samples in profile views to solve the third problem of data labelling. Experiments on the challenging AFLW database show that our approach achieves significant improvements over state-of-the-art methods.",
"In this paper, a new technique for modeling textured 3D faces is introduced. 3D faces can either be generated automatically from one or more photographs, or modeled directly through an intuitive user interface. Users are assisted in two key problems of computer aided face modeling. First, new face images or new 3D face models can be registered automatically by computing dense one-to-one correspondence to an internal face model. Second, the approach regulates the naturalness of modeled faces avoiding faces with an “unlikely” appearance. Starting from an example set of 3D face models, we derive a morphable face model by transforming the shape and texture of the examples into a vector space representation. New faces and expressions can be modeled by forming linear combinations of the prototypes. Shape and texture constraints derived from the statistics of our example faces are used to guide manual modeling or automated matching algorithms. We show 3D face reconstructions from single images and their applications for photo-realistic image manipulations. We also demonstrate face manipulations according to complex parameters such as gender, fullness of a face or its distinctiveness.",
"Face alignment is a classic problem in the computer vision field. Previous works mostly focus on sparse alignment with a limited number of facial landmark points, i.e., facial landmark detection. In this paper, for the first time, we aim at providing a very dense 3D alignment for large-pose face images. To achieve this, we train a CNN to estimate the 3D face shape, which not only aligns limited facial landmarks but also fits face contours and SIFT feature points. Moreover, we also address the bottleneck of training CNN with multiple datasets, due to different landmark markups on different datasets, such as 5, 34, 68. Experimental results show our method not only provides high-quality, dense 3D face fitting but also outperforms the state-of-the-art facial landmark detection methods on challenging datasets. Our model can run at real time during testing and it's available at http: cvlab.cse.msu.edu project-pifa.html.",
"We propose a straightforward method that simultaneously reconstructs the 3D facial structure and provides dense alignment. To achieve this, we design a 2D representation called UV position map which records the 3D shape of a complete face in UV space, then train a simple Convolutional Neural Network to regress it from a single 2D image. We also integrate a weight mask into the loss function during training to improve the performance of the network. Our method does not rely on any prior face model, and can reconstruct full facial geometry along with semantic meaning. Meanwhile, our network is very light-weighted and spends only 9.8 ms to process an image, which is extremely faster than previous works. Experiments on multiple challenging datasets show that our method surpasses other state-of-the-art methods on both reconstruction and alignment tasks by a large margin. Code is available at https: github.com YadiraF PRNet."
]
} |
1904.05562 | 2938318677 | Recently, 3D face reconstruction and face alignment tasks are gradually combined into one task: 3D dense face alignment. Its goal is to reconstruct the 3D geometric structure of face with pose information. In this paper, we propose a graph convolution network to regress 3D face coordinates. Our method directly performs feature learning on the 3D face mesh, where the geometric structure and details are well preserved. Extensive experiments show that our approach gains superior performance over state-of-the-art methods on several challenging datasets. | CNNs are suitable to process grid-like data such as images. By contrast, graph convolution networks (GCNs) are suitable to process graph data such as 3D mesh. A comprehensive overview of GCNs is provided by @cite_25 . @cite_7 define the first graph convolution operator on meshes by parameterizing the surface around each point using geodesic polar coordinates and performing graph convolution on the angular bins. Then, different parameterization methods are proposed by @cite_13 @cite_5 , but the manners of graph convolution are similar. These methods only present generalization of convolutions to meshes. They do not design sampling operation on meshes. Therefore, coarse-to-fine features cannot be captured. | {
"cite_N": [
"@cite_5",
"@cite_13",
"@cite_25",
"@cite_7"
],
"mid": [
"2558460151",
"2963425704",
"2558748708",
"2963021451"
],
"abstract": [
"Deep learning has achieved a remarkable performance breakthrough in several fields, most notably in speech recognition, natural language processing, and computer vision. In particular, convolutional neural network (CNN) architectures currently produce state-of-the-art performance on a variety of image analysis tasks such as object detection and recognition. Most of deep learning research has so far focused on dealing with 1D, 2D, or 3D Euclidean-structured data such as acoustic signals, images, or videos. Recently, there has been an increasing interest in geometric deep learning, attempting to generalize deep learning methods to non-Euclidean structured data such as graphs and manifolds, with a variety of applications from the domains of network analysis, computational social science, or computer graphics. In this paper, we propose a unified framework allowing to generalize CNN architectures to non-Euclidean domains (graphs and manifolds) and learn local, stationary, and compositional task-specific features. We show that various non-Euclidean CNN methods previously proposed in the literature can be considered as particular instances of our framework. We test the proposed method on standard tasks from the realms of image-, graph-and 3D shape analysis and show that it consistently outperforms previous approaches.",
"Convolutional neural networks have achieved extraordinary results in many computer vision and pattern recognition applications; however, their adoption in the computer graphics and geometry processing communities is limited due to the non-Euclidean structure of their data. In this paper, we propose Anisotropic Convolutional Neural Network (ACNN), a generalization of classical CNNs to non-Euclidean domains, where classical convolutions are replaced by projections over a set of oriented anisotropic diffusion kernels. We use ACNNs to effectively learn intrinsic dense correspondences between deformable shapes, a fundamental problem in geometry processing, arising in a wide variety of applications. We tested ACNNs performance in very challenging settings, achieving state-of-the-art results on some of the most difficult recent correspondence benchmarks.",
"Many scientific fields study data with an underlying structure that is non-Euclidean. Some examples include social networks in computational social sciences, sensor networks in communications, functional networks in brain imaging, regulatory networks in genetics, and meshed surfaces in computer graphics. In many applications, such geometric data are large and complex (in the case of social networks, on the scale of billions) and are natural targets for machine-learning techniques. In particular, we would like to use deep neural networks, which have recently proven to be powerful tools for a broad range of problems from computer vision, natural-language processing, and audio analysis. However, these tools have been most successful on data with an underlying Euclidean or grid-like structure and in cases where the invariances of these structures are built into networks used to model them.",
"Feature descriptors play a crucial role in a wide range of geometry analysis and processing applications, including shape correspondence, retrieval, and segmentation. In this paper, we introduce Geodesic Convolutional Neural Networks (GCNN), a generalization of the convolutional neural networks (CNN) paradigm to non-Euclidean manifolds. Our construction is based on a local geodesic system of polar coordinates to extract \"patches\", which are then passed through a cascade of filters and linear and non-linear operators. The coefficients of the filters and linear combination weights are optimization variables that are learned to minimize a task-specific cost function. We use ShapeNet to learn invariant shape features, allowing to achieve state-of-the-art performance in problems such as shape description, retrieval, and correspondence."
]
} |
1904.05562 | 2938318677 | Recently, 3D face reconstruction and face alignment tasks are gradually combined into one task: 3D dense face alignment. Its goal is to reconstruct the 3D geometric structure of face with pose information. In this paper, we propose a graph convolution network to regress 3D face coordinates. Our method directly performs feature learning on the 3D face mesh, where the geometric structure and details are well preserved. Extensive experiments show that our approach gains superior performance over state-of-the-art methods on several challenging datasets. | @cite_4 develops the spectral graph convolution in Fourier space. However, it is computationally expensive and unable to obtain the local features on the graph. To address these problems, ChebyNet @cite_29 formulates spectral convolution as a recursive Chebyshev polynomial, which avoids computing the Fourier basis. Recently, CoMA @cite_14 extends ChebyNet to process 3D meshes. It constructs an autoencoder to learn a latent representation of 3D face and introduces a mesh pooling operator. By utilizing the spectral graph convolution and mesh sampling operations, CoMA @cite_14 obtains state-of-the-art results in 3D face modeling. Motivated by CoMA @cite_14 , for the first time, this work uses graph convolution for 3D dense face alignment. | {
"cite_N": [
"@cite_29",
"@cite_14",
"@cite_4"
],
"mid": [
"2964321699",
"2883221003",
"1662382123"
],
"abstract": [
"In this work, we are interested in generalizing convolutional neural networks (CNNs) from low-dimensional regular grids, where image, video and speech are represented, to high-dimensional irregular domains, such as social networks, brain connectomes or words' embedding, represented by graphs. We present a formulation of CNNs in the context of spectral graph theory, which provides the necessary mathematical background and efficient numerical schemes to design fast localized convolutional filters on graphs. Importantly, the proposed technique offers the same linear computational complexity and constant learning complexity as classical CNNs, while being universal to any graph structure. Experiments on MNIST and 20NEWS demonstrate the ability of this novel deep learning system to learn local, stationary, and compositional features on graphs.",
"Learned 3D representations of human faces are useful for computer vision problems such as 3D face tracking and reconstruction from images, as well as graphics applications such as character generation and animation. Traditional models learn a latent representation of a face using linear subspaces or higher-order tensor generalizations. Due to this linearity, they can not capture extreme deformations and non-linear expressions. To address this, we introduce a versatile model that learns a non-linear representation of a face using spectral convolutions on a mesh surface. We introduce mesh sampling operations that enable a hierarchical mesh representation that captures non-linear variations in shape and expression at multiple scales within the model. In a variational setting, our model samples diverse realistic 3D faces from a multivariate Gaussian distribution. Our training data consists of 20,466 meshes of extreme expressions captured over 12 different subjects. Despite limited training data, our trained model outperforms state-of-the-art face models with 50 lower reconstruction error, while using 75 fewer parameters. We show that, replacing the expression space of an existing state-of-the-art face model with our model, achieves a lower reconstruction error. Our data, model and code are available at http: coma.is.tue.mpg.de .",
"Convolutional Neural Networks are extremely efficient architectures in image and audio recognition tasks, thanks to their ability to exploit the local translational invariance of signal classes over their domain. In this paper we consider possible generalizations of CNNs to signals defined on more general domains without the action of a translation group. In particular, we propose two constructions, one based upon a hierarchical clustering of the domain, and another based on the spectrum of the graph Laplacian. We show through experiments that for low-dimensional graphs it is possible to learn convolutional layers with a number of parameters independent of the input size, resulting in efficient deep architectures."
]
} |
1904.05200 | 2950461368 | Recent years have witnessed the quick progress of the hyperspectral images (HSI) classification. Most of existing studies either heavily rely on the expensive label information using the supervised learning or can hardly exploit the discriminative information borrowed from related domains. To address this issues, in this paper we show a novel framework addressing HSI classification based on the domain adaptation (DA) with active learning (AL). The main idea of our method is to retrain the multi-kernel classifier by utilizing the available labeled samples from source domain, and adding minimum number of the most informative samples with active queries in the target domain. The proposed method adaptively combines multiple kernels, forming a DA classifier that minimizes the bias between the source and target domains. Further equipped with the nested actively updating process, it sequentially expands the training set and gradually converges to a satisfying level of classification performance. We study this active adaptation framework with the Margin Sampling (MS) strategy in the HSI classification task. Our experimental results on two popular HSI datasets demonstrate its effectiveness. | In the literature, support vector machine (SVM) has become one of the most successfully used techniques for hyperspectral image classification @cite_16 , mainly due to the fact that SVM can deal with the high-dimensional and noisy data, with the help of the sparse set of the support vectors @cite_10 and the powerful kernel tricks @cite_25 . This has been proved in many applications like biological problems, which involve high-dimensional, noisy data, for which SVMs are known to behave well compared to other statistical or machine learning methods @cite_8 . Therefore, many kernel-based SVM methods have been studied to capture the complex semantic structure of the hyperspectal images. Basically, these methods first uplift data in the original feature space to a high-dimensional kernel space, and then solve the linear classification problem in the uplifted space. With the supervised information from many labeled samples fed to the model, the kernel-based solutions have demonstrated excellent performance in hyperspectral data classification, in terms of accuracy and robustness @cite_35 @cite_41 @cite_39 @cite_33 @cite_20 . | {
"cite_N": [
"@cite_35",
"@cite_33",
"@cite_8",
"@cite_41",
"@cite_39",
"@cite_16",
"@cite_10",
"@cite_25",
"@cite_20"
],
"mid": [
"2104269704",
"2517660364",
"1580819266",
"2164330327",
"2153409933",
"2136251662",
"2741028958",
"2079202281",
"2547263678"
],
"abstract": [
"This paper presents the framework of kernel-based methods in the context of hyperspectral image classification, illustrating from a general viewpoint the main characteristics of different kernel-based approaches and analyzing their properties in the hyperspectral domain. In particular, we assess performance of regularized radial basis function neural networks (Reg-RBFNN), standard support vector machines (SVMs), kernel Fisher discriminant (KFD) analysis, and regularized AdaBoost (Reg-AB). The novelty of this work consists in: 1) introducing Reg-RBFNN and Reg-AB for hyperspectral image classification; 2) comparing kernel-based methods by taking into account the peculiarities of hyperspectral images; and 3) clarifying their theoretical relationships. To these purposes, we focus on the accuracy of methods when working in noisy environments, high input dimension, and limited training sets. In addition, some other important issues are discussed, such as the sparsity of the solutions, the computational burden, and the capability of the methods to provide outputs that can be directly interpreted as probabilities.",
"In recent years, many studies on hyperspectral image classification have shown that using multiple features can effectively improve the classification accuracy. As a very powerful means of learning, multiple kernel learning (MKL) can conveniently be embedded in a variety of characteristics. This paper proposes a class-specific sparse MKL (CS-SMKL) framework to improve the capability of hyperspectral image classification. In terms of the features, extended multiattribute profiles are adopted because it can effectively represent the spatial and spectral information of hyperspectral images. CS-SMKL classifies the hyperspectral images, simultaneously learns class-specific significant features, and selects class-specific weights. Using an @math -norm constraint (i.e., group lasso) as the regularizer, we can enforce the sparsity at the group feature level and automatically learn a compact feature set for the classification of any two classes. More precisely, our CS-SMKL determines the associated weights of optimal base kernels for any two classes and results in improved classification performances. The advantage of the proposed method is that only the features useful for the classification of any two classes can be retained, which leads to greatly enhanced discriminability. Experiments are conducted on three hyperspectral data sets. The experimental results show that the proposed method achieves better performances for hyperspectral image classification compared with several state-of-the-art algorithms, and the results confirm the capability of the method in selecting the useful features.",
"Bioinformatics, a field devoted to the interpretation and analysis of biological data using computational techniques, has evolved tremendously in recent years due to the explosive growth of biological information generated by the scientific community. Soft computing is a consortium of methodologies that work synergistically and provides, in one form or another, flexible information processing capabilities for handling real-life ambiguous situations. Several research articles dealing with the application of soft computing tools to bioinformatics have been published in the recent past; however, they are scattered in different journals, conference proceedings and technical reports, thus causing inconvenience to readers, students and researchers. This book, unique in its nature, is aimed at providing a treatise in a unified framework, with both theoretical and experimental results, describing the basic principles of soft computing and demonstrating the various ways in which they can be used for analyzing biological data in an efficient manner. Interesting research articles from eminent scientists around the world are brought together in a systematic way such that the reader will be able to understand the issues and challenges in this domain, the existing ways of tackling them, recent trends, and future directions. This book is the first of its kind to bring together two important research areas, soft computing and bioinformatics, in order to demonstrate how the tools and techniques in the former can be used for efficiently solving several problems in the latter.",
"This letter presents a framework of composite kernel machines for enhanced classification of hyperspectral images. This novel method exploits the properties of Mercer's kernels to construct a family of composite kernels that easily combine spatial and spectral information. This framework of composite kernels demonstrates: 1) enhanced classification accuracy as compared to traditional approaches that take into account the spectral information only: 2) flexibility to balance between the spatial and spectral information in the classifier; and 3) computational efficiency. In addition, the proposed family of kernel classifiers opens a wide field for future developments in which spatial and spectral information can be easily integrated.",
"This paper presents a semi-supervised graph-based method for the classification of hyperspectral images. The method is designed to handle the special characteristics of hyperspectral images, namely, high-input dimension of pixels, low number of labeled samples, and spatial variability of the spectral signature. To alleviate these problems, the method incorporates three ingredients, respectively. First, being a kernel-based method, it combats the curse of dimensionality efficiently. Second, following a semi-supervised approach, it exploits the wealth of unlabeled samples in the image, and naturally gives relative importance to the labeled ones through a graph-based methodology. Finally, it incorporates contextual information through a full family of composite kernels. Noting that the graph method relies on inverting a huge kernel matrix formed by both labeled and unlabeled samples, we originally introduce the Nystro umlm method in the formulation to speed up the classification process. The presented semi-supervised-graph-based method is compared to state-of-the-art support vector machines in the classification of hyperspectral data. The proposed method produces better classification maps, which capture the intrinsic structure collectively revealed by labeled and unlabeled points. Good and stable accuracy is produced in ill-posed classification problems (high dimensional spaces and low number of labeled samples). In addition, the introduction of the composite-kernel framework drastically improves results, and the new fast formulation ranks almost linearly in the computational cost, rather than cubic as in the original method, thus allowing the use of this method in remote-sensing applications.",
"This paper addresses the problem of the classification of hyperspectral remote sensing images by support vector machines (SVMs). First, we propose a theoretical discussion and experimental analysis aimed at understanding and assessing the potentialities of SVM classifiers in hyperdimensional feature spaces. Then, we assess the effectiveness of SVMs with respect to conventional feature-reduction-based approaches and their performances in hypersubspaces of various dimensionalities. To sustain such an analysis, the performances of SVMs are compared with those of two other nonparametric classifiers (i.e., radial basis function neural networks and the K-nearest neighbor classifier). Finally, we study the potentially critical issue of applying binary SVMs to multiclass problems in hyperspectral data. In particular, four different multiclass strategies are analyzed and compared: the one-against-all, the one-against-one, and two hierarchical tree-based strategies. Different performance indicators have been used to support our experimental studies in a detailed and accurate way, i.e., the classification accuracy, the computational time, the stability to parameter setting, and the complexity of the multiclass architecture. The results obtained on a real Airborne Visible Infrared Imaging Spectroradiometer hyperspectral dataset allow to conclude that, whatever the multiclass strategy adopted, SVMs are a valid and effective alternative to conventional pattern recognition approaches (feature-reduction procedures combined with a classification method) for the classification of hyperspectral remote sensing data.",
"",
"Recently hashing has become attractive in large-scale visual search, owing to its theoretical guarantee and practical success. However, most of the state-of-the-art hashing methods can only employ a single feature type to learn hashing functions. Related research on image search, clustering, and other domains has proved the advantages of fusing multiple features. In this paper we propose a novel multiple feature kernel hashing framework, where hashing functions are learned to preserve certain similarities with linearly combined multiple kernels corresponding to different features. The framework is not only compatible with general types of data and diverse types of similarities indicated by different visual features, but also general for both supervised and unsupervised scenarios. We present efficient alternating optimization algorithms to learn both the hashing functions and the optimal kernel combination. Experimental results on three large-scale benchmarks CIFAR-10, NUS-WIDE and a-TRECVID show that the proposed approach can achieve superior accuracy and efficiency over state-of-the-art methods. HighlightsWe propose a generic multiple feature hashing framework using multiple kernels.Visual features are implicitly mapped and concatenated to reduce complexity.We formulate both supervised and unsupervised hashing problems in the framework.Alternating optimization ways efficiently learn hashing functions and the kernel space.Experiments validate the superior performances and efficiency of the proposed approach.",
"Hyperspectral images play an important role in real-world applications, such as recognition and remote sensing, etc. How to enhance the spatial resolution of hyperspectral image is still a challenging problem in this field. In this paper, we propose a novel hyperspectral image super-resolution approach by jointly incorporating the sparse, low-rank constraints and spectral mixture priori into a linear unmixing framework, which will make the unmixing framework more consistent with the real-world scenarios of the spectral mixture. Experiments on two public databases show that our proposed approach achieves much lower average reconstruction errors than other state-of-the-art methods."
]
} |
1904.05200 | 2950461368 | Recent years have witnessed the quick progress of the hyperspectral images (HSI) classification. Most of existing studies either heavily rely on the expensive label information using the supervised learning or can hardly exploit the discriminative information borrowed from related domains. To address this issues, in this paper we show a novel framework addressing HSI classification based on the domain adaptation (DA) with active learning (AL). The main idea of our method is to retrain the multi-kernel classifier by utilizing the available labeled samples from source domain, and adding minimum number of the most informative samples with active queries in the target domain. The proposed method adaptively combines multiple kernels, forming a DA classifier that minimizes the bias between the source and target domains. Further equipped with the nested actively updating process, it sequentially expands the training set and gradually converges to a satisfying level of classification performance. We study this active adaptation framework with the Margin Sampling (MS) strategy in the HSI classification task. Our experimental results on two popular HSI datasets demonstrate its effectiveness. | There are several ways in DA research to migrate the classifiers among the source and the target domains. One of the typical strategies is to make the data distributions more similar across the domains to train a single model that can simultaneously classify the source and target domains. @cite_24 designed a multiple-kernel DA technique to learn a discriminative model by simultaneously minimizing both the SVM structural risk and the distribution divergence. @cite_42 adapted the maximum-likelihood classifier to the target domain through updating the classifier parameters. Even with the promising progress achieved by the DA based HSI classification methods, it is still beyond the desired performance without the supervised information from target domain, mainly due to the divergence among the source and the target domains. | {
"cite_N": [
"@cite_24",
"@cite_42"
],
"mid": [
"2036842392",
"2137900926"
],
"abstract": [
"This letter presents a novel semisupervised method for addressing a domain adaptation problem in the classification of hyperspectral data. To overcome the influence of distribution bias between the source and target domains, we introduce the domain transfer multiple-kernel learning to simultaneously minimize the maximum mean discrepancy criterion and the structural risk functional of support vector machines. Then, the pairwise binary classifiers are merged as the multiclass classifier for solving the classification problem in hyperspectral data. Both bias and nonbias sampling strategies are introduced to evaluate the robustness of the proposed method against the spectral distribution bias. The results obtained from real data sets show that the proposed method can achieve higher classification accuracy even with cross-domain distribution bias and provide robust solutions with different labeled and unlabeled data sizes.",
"An unsupervised retraining technique for a maximum likelihood (ML) classifier is presented. The proposed technique allows the classifier's parameters, obtained by supervised learning on a specific image, to be updated in a totally unsupervised way on the basis of the distribution of a new image to be classified. This enables the classifier to provide a high accuracy for the new image even when the corresponding training set is not available."
]
} |
1904.05200 | 2950461368 | Recent years have witnessed the quick progress of the hyperspectral images (HSI) classification. Most of existing studies either heavily rely on the expensive label information using the supervised learning or can hardly exploit the discriminative information borrowed from related domains. To address this issues, in this paper we show a novel framework addressing HSI classification based on the domain adaptation (DA) with active learning (AL). The main idea of our method is to retrain the multi-kernel classifier by utilizing the available labeled samples from source domain, and adding minimum number of the most informative samples with active queries in the target domain. The proposed method adaptively combines multiple kernels, forming a DA classifier that minimizes the bias between the source and target domains. Further equipped with the nested actively updating process, it sequentially expands the training set and gradually converges to a satisfying level of classification performance. We study this active adaptation framework with the Margin Sampling (MS) strategy in the HSI classification task. Our experimental results on two popular HSI datasets demonstrate its effectiveness. | Since the labeled information in target domain can directly and notably improve the classification performance, it is obviously helpful that we can exploit a small number of labeled data to boost the classification performance, which at the same time only brings a quite limited additional cost for the labeled data collection. The active learning (AL) strategies have been widely studied in the literature to tackle such a challenging task in recent years @cite_40 @cite_12 @cite_46 @cite_31 , with the aim to exploit the information available from unlabeled data and to enrich the labeled data. In AL process, labelling originally unlabeled data is usually completed by a user according to an specific informative measure. | {
"cite_N": [
"@cite_46",
"@cite_40",
"@cite_31",
"@cite_12"
],
"mid": [
"2118834604",
"2426031434",
"2444115989",
"2157575532"
],
"abstract": [
"Semi-supervised learning and active learning are important techniques to build more accurate model while labeled data are scarce. The objective of this paper is combining both to effectively relieve user labor for multi-class annotation. We propose a novel graph-based active semi-supervised learning framework which aim at efficiently learning a multi-class model with minimal human labor. In particular, we propose Minimize Expected Global Uncertainty algorithm to actively select examples (for labels), which naturally integrates with the probabilistic results of graph-based semi-supervised learning. Meanwhile, we update the model incrementally by decomposed formulation while the new example are incorporated for training, which only has the time complexity of O(n), compared to the original re-training of O(n 3 ). Extensive evaluations over three real-world datasets demonstrate that our proposed method has the superior performance comparing with the baselines and the capability to efficiently build more accurate model with fractional human labor.",
"Support vector machines have met with significant success in numerous real-world learning tasks. However, like most machine learning algorithms, they are generally applied using a randomly selected training set classified in advance. In many settings, we also have the option of using pool-based active learning. Instead of using a randomly selected training set, the learner has access to a pool of unlabeled instances and can request the labels for some number of them. We introduce a new algorithm for performing active learning with support vector machines, i.e., an algorithm for choosing which instances to request next. We provide a theoretical motivation for the algorithm using the notion of a version space. We present experimental results showing that employing our active learning method can significantly reduce the need for labeled training instances in both the standard inductive and transductive settings.",
"Hashing has become an increasingly popular technique for fast nearest neighbor search. Despite its successful progress in classic pointto-point search, there are few studies regarding point-to-hyperplane search, which has strong practical capabilities of scaling up applications like active learning with SVMs. Existing hyperplane hashing methods enable the fast search based on randomly generated hash codes, but still suffer from a low collision probability and thus usually require long codes for a satisfying performance. To overcome this problem, this paper proposes a multilinear hyperplane hashing that generates a hash bit using multiple linear projections. Our theoretical analysis shows that with an even number of random linear projections, the multilinear hash function possesses strong locality sensitivity to hyperplane queries. To leverage its sensitivity to the angle distance, we further introduce an angular quantization based learning framework for compact multilinear hashing, which considerably boosts the search performance with less hash bits. Experiments with applications to large-scale (up to one million) active learning on two datasets demonstrate the overall superiority of the proposed approach.",
"We consider the problem of retrieving the database points nearest to a given hyperplane query without exhaustively scanning the entire database. For this problem, we propose two hashing-based solutions. Our first approach maps the data to 2-bit binary keys that are locality sensitive for the angle between the hyperplane normal and a database point. Our second approach embeds the data into a vector space where the euclidean norm reflects the desired distance between the original points and hyperplane query. Both use hashing to retrieve near points in sublinear time. Our first method's preprocessing stage is more efficient, while the second has stronger accuracy guarantees. We apply both to pool-based active learning: Taking the current hyperplane classifier as a query, our algorithm identifies those points (approximately) satisfying the well-known minimal distance-to-hyperplane selection criterion. We empirically demonstrate our methods' tradeoffs and show that they make it practical to perform active selection with millions of unlabeled points."
]
} |
1904.05200 | 2950461368 | Recent years have witnessed the quick progress of the hyperspectral images (HSI) classification. Most of existing studies either heavily rely on the expensive label information using the supervised learning or can hardly exploit the discriminative information borrowed from related domains. To address this issues, in this paper we show a novel framework addressing HSI classification based on the domain adaptation (DA) with active learning (AL). The main idea of our method is to retrain the multi-kernel classifier by utilizing the available labeled samples from source domain, and adding minimum number of the most informative samples with active queries in the target domain. The proposed method adaptively combines multiple kernels, forming a DA classifier that minimizes the bias between the source and target domains. Further equipped with the nested actively updating process, it sequentially expands the training set and gradually converges to a satisfying level of classification performance. We study this active adaptation framework with the Margin Sampling (MS) strategy in the HSI classification task. Our experimental results on two popular HSI datasets demonstrate its effectiveness. | Due to the promising performance, many efforts have been devoted to the integration of AL in domain adaptation based HSI classification @cite_48 @cite_3 @cite_37 @cite_22 @cite_23 . @cite_3 offered an iterative AL process by adding the most informative samples to the training set, while removing the source-domain samples that do not fit with the distributions of the classes in the target domain. @cite_22 is a framework that efficiently combines the DA and AL techniques, and the most informative pixels are sampled with active queries from the target image while adapting the obtained classifier using a transfer learning strategy. In @cite_37 , a DA framework equipped with the active selection pursued the training samples in unknown areas using the strategies based on uncertainty and clustering. These active learning methods are able to alleviate the expensive cost on the collection of labeled data by interactively generating the labeled data. | {
"cite_N": [
"@cite_37",
"@cite_22",
"@cite_48",
"@cite_3",
"@cite_23"
],
"mid": [
"2154772499",
"2072243389",
"2141039193",
"2093596219",
"2168795822"
],
"abstract": [
"The validity of training samples collected in field campaigns is crucial for the success of land use classification models. However, such samples often suffer from a sample selection bias and do not represent the variability of spectra that can be encountered in the entire image. Therefore, to maximize classification performance, one must perform adaptation of the first model to the new data distribution. In this paper, we propose to perform adaptation by sampling new training examples in unknown areas of the image. Our goal is to select these pixels in an intelligent fashion that minimizes their number and maximizes their information content. Two strategies based on uncertainty and clustering of the data space are considered to perform active selection. Experiments on urban and agricultural images show the great potential of the proposed strategy to perform model adaptation.",
"We propose a procedure that efficiently adapts a classifier trained on a source image to a target image with similar spectral properties. The adaptation is carried out by adding new relevant training samples with active queries in the target domain following a strategy specifically designed for the case where class distributions have shifted between the two acquisitions. In fact, the procedure consists of two nested algorithms. An active selection of the pixels to be labeled is performed on a set of candidates of the target image in order to select the most informative pixels. Along the inclusion of the pixels to the training set, the weights associated with these samples are iteratively updated using different criteria, depending on their origin (source or target image). We study this adaptation framework in combination with a SVM classifier accepting instance weights. Experiments on two VHR QuickBird images and on a hyperspectral AVIRIS image prove the validity of the proposed adaptive approach with respect to existing techniques not involving any adjustments to the target domain.",
"We present a novel technique for addressing domain adaptation problems in the classification of remote sensing images with active learning. Domain adaptation is the important problem of adapting a supervised classifier trained on a given image (source domain) to the classification of another similar (but not identical) image (target domain) acquired on a different area, or on the same area at a different time. The main idea of the proposed approach is to iteratively labeling and adding to the training set the minimum number of the most informative samples from target domain, while removing the source-domain samples that does not fit with the distributions of the classes in the target domain. In this way, the classification system exploits already available information, i.e., the labeled samples of source domain, in order to minimize the number of target domain samples to be labeled, thus reducing the cost associated to the definition of the training set for the classification of the target domain. Experimental results obtained in the classification of a hyperspectral image confirm the effectiveness of the proposed technique.",
"This paper presents a novel technique for addressing domain adaptation (DA) problems with active learning (AL) in the classification of remote sensing images. DA models the important problem of adapting a supervised classifier trained on a given image (source domain) to the classification of another similar but not identical image (target domain) acquired on a different area. The main idea of the proposed approach is iteratively labeling and adding to the training set the minimum number of the most informative samples from the target domain, while removing the source-domain samples that do not fit with the distributions of the classes in the target domain. In this way, the classification system exploits already available information, i.e., the labeled samples of source domain, in order to minimize the number of target domain samples to be labeled, thus reducing the cost associated to the definition of the training set for the classification of the target domain. In addition, we define a convergence criterion that allows the technique to stop the iterative AL process on the target domain without relying on the availability of a test set for it. This is an important contribution, as in operational applications, it is not realistic to assume that a test set for the target domain is available. Experimental results obtained in the classification of very high resolution and hyperspectral images confirm the effectiveness of the proposed technique.",
"In remote sensing image classification, it is commonly assumed that the distribution of the classes is stable over the entire image. This way, training pixels labeled by photointerpretation are assumed to be representative of the whole image. However, differences in distribution of the classes throughout the image make this assumption weak and a model built on a single area may be suboptimal when applied to the rest of the image. In this paper, we investigate the use of active learning to correct the shifts that may appear when training and test data do not come from the same distribution. Experiments are carried out on a VHR remote sensing classification scenario showing that active learning can effectively learn the covariance shift and provide robust solutions."
]
} |
1904.05351 | 2935891278 | Neural networks based vocoders have recently demonstrated the powerful ability to synthesize high quality speech. These models usually generate samples by conditioning on some spectrum features, such as Mel-spectrum. However, these features are extracted by using speech analysis module including some processing based on the human knowledge. In this work, we proposed RawNet, a truly end-to-end neural vocoder, which use a coder network to learn the higher representation of signal, and an autoregressive voder network to generate speech sample by sample. The coder and voder together act like an auto-encoder network, and could be jointly trained directly on raw waveform without any human-designed features. The experiments on the Copy-Synthesis tasks show that RawNet can achieve the comparative synthesized speech quality with LPCNet, with a smaller model architecture and faster speech generation at the inference step. | The architecture of RawNet is like to an autoencoder model, in which the encoder and decoder networks could correspond the coder and voder network of RawNet. There are some researches in employing an auto-encoder for extracting relative parameters for speech synthesis task. For example, @cite_20 @cite_12 use an autoencoder to extract excitation parameters, which is required by a traditional vocoder. In @cite_14 , an autoencoder based, non-linear and data-driven method is used to extract low-dimensional features from the FFT spectral envelop, instead of using the speech analysis module based on human knowledge. @cite_14 also concludes that the proposed model outperforms the one which is based on the conventional feature extraction. The difference of RawNet to the above mentioned methods is that, RawNet directly takes waveform samples as input instead of just treats autoencoder as a feature dimension reduction method. | {
"cite_N": [
"@cite_14",
"@cite_12",
"@cite_20"
],
"mid": [
"2395849284",
"2398742733",
"2153750031"
],
"abstract": [
"In the state-of-the-art statistical parametric speech synthesis system, a speech analysis module, e.g. STRAIGHT spectral analysis, is generally used for obtaining accurate and stable spectral envelopes, and then low-dimensional acoustic features extracted from obtained spectral envelopes are used for training acoustic models. However, a spectral envelope estimation algorithm used in such a speech analysis module includes various processing derived from human knowledge. In this paper, we present our investigation of deep autoencoder based, non-linear, data-driven and unsupervised low-dimensional feature extraction using FFT spectral envelopes for statistical parametric speech synthesis. Experimental results showed that a text-to-speech synthesis system using deep auto-encoder based low-dimensional feature extraction from FFT spectral envelopes is indeed a promising approach.",
"This paper studies a deep neural network (DNN) based voice source modelling method in the synthesis of speech with varying vocal effort. The new trainable voice source model learns a mapping between the acoustic features and the time-domain pitch-synchronous glottal flow waveform using a DNN. The voice source model is trained with various speech material from breathy, normal, and Lombard speech. In synthesis, a normal voice is first adapted to a desired style, and using the flexible DNN-based voice source model, a style-specific excitation waveform is automatically generated based on the adapted acoustic features. The proposed voice source model is compared to a robust and high-quality excitation modelling method based on manually selected mean glottal flow pulses for each vocal effort level and using a spectral matching filter to correctly match the voice source spectrum to a desired style. Subjective evaluations show that the proposed DNN-based method is rated comparable to the baseline method, but avoids the manual selection of the pulses and is computationally faster than a system using a spectral matching filter.",
"HMM-TTS synthesis is a popular approach toward flexible, low-footprint, data driven systems that produce highly intelligible speech. In spite of these strengths, speech generated by these systems exhibit some degradation in quality, attributable to an inadequacy in modeling the excitation signal that drives the parametric models of the vocal tract. This paper proposes a novel method for modeling the excitation as a low-dimensional set of coefficients, based on a non-linear map learned through an autoencoder. Through analysis-and-resynthesis experiments, and a formal listening test, we show that this model produces speech of higher perceptual quality compared to conventional pulse-excited speech signals at the p ≪ 0.01 significance level."
]
} |
1904.05335 | 2782416476 | Stochastic block models (SBMs) have been playing an important role in modeling clusters or community structures of network data. But, it is incapable of handling several complex features ubiquitously exhibited in real-world networks, one of which is the power-law degree characteristic. To this end, we propose a new variant of SBM, termed power-law degree SBM (PLD-SBM), by introducing degree decay variables to explicitly encode the varying degree distribution over all nodes. With an exponential prior, it is proved that PLD-SBM approximately preserves the scale-free feature in real networks. In addition, from the inference of variational E-Step, PLD-SBM is indeed to correct the bias inherited in SBM with the introduced degree decay factors. Furthermore, experiments conducted on both synthetic networks and two real-world datasets including Adolescent Health Data and the political blogs network verify the effectiveness of the proposed model in terms of cluster prediction accuracies. | A stochastic block model (SBM) handles the task of recovering blocks, groups or community structures in networks @cite_13 @cite_6 . | {
"cite_N": [
"@cite_13",
"@cite_6"
],
"mid": [
"2132589243",
"2315775489"
],
"abstract": [
"Mining patterns in graphs has become an important issue in real applications, such as bioinformatics and web mining. We address a graph clustering problem where a cluster is a set of densely connected nodes, under a practical setting that 1) the input is multiple graphs which share a set of nodes but have different edges and 2) a true cluster cannot be found in all given graphs. For this problem, we propose a probabilistic generative model and a robust learning scheme based on variational Bayesian estimation. A key feature of our probabilistic framework is that not only nodes but also given graphs can be clustered at the same time, allowing our model to capture clusters found in only part of all given graphs. We empirically evaluated the effectiveness of the proposed framework on not only a variety of synthetic graphs but also real gene networks, demonstrating that our proposed approach can improve the clustering performance of competing methods in both synthetic and real data.",
"Distributed and dynamic networks are ubiquitous in many real-world applications. Due to the huge-scale, decentralized, and dynamic characteristics, the global topological view is either too hard to obtain or even not available. So, most existing community detection methods working on the global view fail to handle such decentralized and dynamic large networks. In this paper, we propose a novel autonomy-oriented computing-based method for community mining (AOCCM) from the multiagent perspective in the distributed environment. In particular, AOCCM utilizes reactive agents to pick the neighborhood node with the largest structural similarity as the candidate node, and thus determine whether it should be added into local community based on the modularity gain. We further improve AOCCM to a more efficient incremental version named AOCCM-i for mining communities from dynamic networks. AOCCM and AOCCM-i can be easily expanded to detect both nonoverlapping and overlapping global community structures. Experimental results on real-life networks demonstrate that the proposed methods can reduce the computational cost by avoiding repeated structural similarity calculation and can still obtain the high-quality communities."
]
} |
1904.05335 | 2782416476 | Stochastic block models (SBMs) have been playing an important role in modeling clusters or community structures of network data. But, it is incapable of handling several complex features ubiquitously exhibited in real-world networks, one of which is the power-law degree characteristic. To this end, we propose a new variant of SBM, termed power-law degree SBM (PLD-SBM), by introducing degree decay variables to explicitly encode the varying degree distribution over all nodes. With an exponential prior, it is proved that PLD-SBM approximately preserves the scale-free feature in real networks. In addition, from the inference of variational E-Step, PLD-SBM is indeed to correct the bias inherited in SBM with the introduced degree decay factors. Furthermore, experiments conducted on both synthetic networks and two real-world datasets including Adolescent Health Data and the political blogs network verify the effectiveness of the proposed model in terms of cluster prediction accuracies. | There is also other literature on addressing degree heterogeneity in networks using spectral methods, for which we leave the references @cite_44 @cite_60 @cite_63 @cite_54 for interested readers. | {
"cite_N": [
"@cite_44",
"@cite_54",
"@cite_63",
"@cite_60"
],
"mid": [
"1524273267",
"2154624311",
"2127475907",
"857387926"
],
"abstract": [
"Understanding a complex network's structure holds the key to understanding its function. The physics community has contributed a multitude of methods and analyses to this cross-disciplinary endeavor. Structural features exist on both the microscopic level, resulting from differences between single node properties, and the mesoscopic level resulting from properties shared by groups of nodes. Disentangling the determinants of network structure on these different scales has remained a major, and so far unsolved, challenge. Here we show how multiscale generative probabilistic exponential random graph models combined with efficient, distributive message-passing inference techniques can be used to achieve this separation of scales, leading to improved detection accuracy of latent classes as demonstrated on benchmark problems. It sheds new light on the statistical significance of motif-distributions in neural networks and improves the link-prediction accuracy as exemplified for gene-disease associations in the highly consequential Online Mendelian Inheritance in Man database.",
"Traditional image representations are not suited to conventional classification methods such as the linear discriminant analysis (LDA) because of the undersample problem (USP): the dimensionality of the feature space is much higher than the number of training samples. Motivated by the successes of the two-dimensional LDA (2DLDA) for face recognition, we develop a general tensor discriminant analysis (GTDA) as a preprocessing step for LDA. The benefits of GTDA, compared with existing preprocessing methods such as the principal components analysis (PCA) and 2DLDA, include the following: 1) the USP is reduced in subsequent classification by, for example, LDA, 2) the discriminative information in the training tensors is preserved, and 3) GTDA provides stable recognition rates because the alternating projection optimization algorithm to obtain a solution of GTDA converges, whereas that of 2DLDA does not. We use human gait recognition to validate the proposed GTDA. The averaged gait images are utilized for gait representation. Given the popularity of Gabor-function-based image decompositions for image understanding and object recognition, we develop three different Gabor-function-based image representations: 1) GaborD is the sum of Gabor filter responses over directions, 2) GaborS is the sum of Gabor filter responses over scales, and 3) GaborSD is the sum of Gabor filter responses over scales and directions. The GaborD, GaborS, and GaborSD representations are applied to the problem of recognizing people from their averaged gait images. A large number of experiments were carried out to evaluate the effectiveness (recognition rate) of gait recognition based on first obtaining a Gabor, GaborD, GaborS, or GaborSD image representation, then using GDTA to extract features and, finally, using LDA for classification. The proposed methods achieved good performance for gait recognition based on image sequences from the University of South Florida (USF) HumanID Database. Experimental comparisons are made with nine state-of-the-art classification methods in gait recognition.",
"We extend spectral methods to random graphs with skewed degree distributions through a degree based normalization closely connected to the normalized Laplacian. The normalization is based on intuition drawn from perturbation theory of random matrices, and has the effect of boosting the expectation of the random adjacency matrix without increasing the variances of its entries, leading to better perturbation bounds. The primary implication of this result lies in the realm of spectral analysis of random graphs with skewed degree distributions, such as the ubiquitous \"power law graphs\". Mihail and Papadimitriou (2002) argued that for randomly generated graphs satisfying a power law degree distribution, spectral analysis of the adjacency matrix simply produces the neighborhoods of the high degree nodes as its eigenvectors, and thus miss any embedded structure. We present a generalization of their model, incorporating latent structure, and prove that after applying our transformation, spectral analysis succeeds in recovering the latent structure with high probability.",
"In this paper, we examine a spectral clustering algorithm for similarity graphs drawn from a simple random graph model, where nodes are allowed to have varying degrees, and we provide theoretical bounds on its performance. The random graph model we study is the Extended Planted Partition (EPP) model, a variant of the classical planted partition model. The standard approach to spectral clustering of graphs is to compute the bottom k singular vectors or eigenvectors of a suitable graph Laplacian, project the nodes of the graph onto these vectors, and then use an iterative clustering algorithm on the projected nodes. However a challenge with applying this approach to graphs generated from the EPP model is that unnormalized Laplacians do not work, and normalized Laplacians do not concentrate well when the graph has a number of low degree nodes. We resolve this issue by introducing the notion of a degree-corrected graph Laplacian. For graphs with many low degree nodes, degree correction has a regularizing eect on the Laplacian. Our spectral clustering algorithm projects the nodes in the graph onto the bottom k right singular vectors of the degree-corrected random-walk Laplacian, and clusters the nodes in this subspace. We show guarantees on the performance of this algorithm, demonstrating that it outputs the correct partition under a wide range of parameter values. Unlike some previous work, our algorithm does not require access to any generative parameters of the model."
]
} |
1904.05213 | 2938403409 | Consider an interference channel consisting of @math transmitters and @math receivers with AWGN noise and complex channel gains, and with @math files in the system. The one-shot @math for this channel is the maximum number of receivers which can be served simultaneously with vanishing probability of error as the @math grows large, under a class of schemes known as schemes. Consider that there exists transmitter and receiver side caches which can store fractions @math and @math of the library respectively. Recent work for this cache-aided interference channel setup shows that, using a carefully designed prefetching(caching) phase, and a one-shot coded delivery scheme combined with a proper choice of beamforming coefficients at the transmitters, we can achieve a @math of @math , where @math and @math which was shown to be almost optimal. The existing scheme involves splitting the file into @math subfiles (the parameter @math is called the ), where @math can be extremely large (in fact, with constant cache fractions, it becomes exponential in @math , for large @math ). In this work, our first contribution is a scheme which achieves the same @math of @math with a smaller subpacketization than prior schemes. Our second contribution is a new coded caching scheme for the interference channel based on projective geometries over finite fields which achieves a one-shot @math of @math , with a subpacketization @math (for some prime power @math ) that is in @math , for small constant cache fraction at the receivers. To the best of our knowledge, this is the first coded caching scheme with subpacketization subexponential in the number of receivers for this setting. | The first work which considered applying the strategy of coded caching to interference channels was @cite_10 . The authors of @cite_10 focused on the case with only transmitter caches, which were exploited to get both interference cancellation and interference alignment gains. This was further expanded upon by @cite_1 where the DoF region of the interference channel with both transmitter and receiver channels was characterized and a near-optimal scheme (optimal except for a multiplicative gap) was obtained using general delivery schemes (beyond one-shot schemes). Coded caching for multi-antenna interference channels was considered in @cite_11 . A one-shot delivery based coded caching scheme for the setup of @cite_8 was presented in @cite_13 , which achieves smaller subpacketization than the scheme in @cite_8 . However the subpacketization still remains exponential in the number of receivers @math . | {
"cite_N": [
"@cite_13",
"@cite_8",
"@cite_1",
"@cite_10",
"@cite_11"
],
"mid": [
"2963935351",
"2589022566",
"2409479574",
"2963360080",
""
],
"abstract": [
"We consider a cache-aided interference network which consists of a library of N files, K<sub>T</sub> transmitters and K<sub>R</sub> receivers (users), each equipped with a local cache of size M<sub>T</sub> and M<sub>R</sub> files respectively, and connected via a discrete-time additive white Gaussian noise channel. Each receiver requests an arbitrary file from the library. The objective is to design a cache placement without knowing the receivers' requests and a communication scheme such that the sum Degrees of Freedom (sum-DoF) of the delivery is maximized. This network model has been investigated by , who proposed a prefetching and a delivery scheme that achieve a sum-DoF of min M<sub>T</sub> K<sub>T</sub> + K<sub>R</sub> M<sub>R</sub>/N, K<sub>R</sub> . One of the biggest limitations of this scheme is the requirement of high subpacketization level. This paper attempts to design new algorithms to reduce the file subpacketization in such a network. In particular, we propose a new approach for both prefetching and linear delivery based on a combinatorial design called hypercube. We show that the required number of packets per file can be exponentially reduced compared to the state-of-the-art scheme proposed by , or the NMA scheme. When M<sub>T</sub> K<sub>T</sub> + K<sub>R</sub> M<sub>R</sub> ≤ K<sub>R</sub>, the achievable one-shot sum-DoF using this approach is M<sub>T</sub> K<sub>T</sub> + K<sub>R</sub> M<sub>R</sub>/N, which shows that 1) the one-shot sum-DoF scales linearly with the aggregate cache size in the network and 2) it is within a factor of 2 to the information-theoretic optimum. Surprisingly, the identical and near optimal sum-DoF performance can be achieved using the hypercube approach with a much less file subpacketization.",
"We consider a system, comprising a library of @math files (e.g., movies) and a wireless network with a @math transmitters, each equipped with a local cache of size of @math files and a @math receivers, each equipped with a local cache of size of @math files. Each receiver will ask for one of the @math files in the library, which needs to be delivered. The objective is to design the cache placement (without prior knowledge of receivers’ future requests) and the communication scheme to maximize the throughput of the delivery. In this setting, we show that the sum degrees-of-freedom (sum-DoF) of @math is achievable, and this is within a factor of 2 of the optimum, under uncoded prefetching and one-shot linear delivery schemes. This result shows that (i) the one-shot sum-DoF scales linearly with the aggregate cache size in the network (i.e., the cumulative memory available at all nodes ), (ii) the transmitters’ caches and receivers’ caches contribute equally in the one-shot sum-DoF, and (iii) caching can offer a throughput gain that scales linearly with the size of the network. To prove the result, we propose an achievable scheme that exploits the redundancy of the content at transmitter’s caches to cooperatively zero-force some outgoing interference, and availability of the unintended content at the receiver’s caches to cancel (subtract) some of the incoming interference. We develop a particular pattern for cache placement that maximizes the overall gains of cache-aided transmit and receive interference cancellations. For the converse, we present an integer optimization problem which minimizes the number of communication blocks needed to deliver any set of requested files to the receivers. We then provide a lower bound on the value of this optimization problem, hence leading to an upper bound on the linear one-shot sum-DoF of the network, which is within a factor of 2 of the achievable sum-DoF.",
"We study the role of caches in wireless interference networks. We focus on content caching and delivery across a Gaussian interference network, where both transmitters and receivers are equipped with caches. We provide a constant-factor approximation of the system's degrees of freedom (DoF), for arbitrary number of transmitters, number of receivers, content library size, receiver cache size, and transmitter cache size (as long as the transmitters combined can store the entire content library among them). We demonstrate approximate optimality with respect to information-theoretic bounds that do not impose any restrictions on the caching and delivery strategies. Our characterization reveals three key insights. First, the approximate DoF is achieved using a strategy that separates the physical and network layers. This separation architecture is thus approximately optimal. Second, we show that increasing transmitter cache memory beyond what is needed to exactly store the entire library between all transmitters does not provide more than a constant-factor benefit to the DoF. A consequence is that transmit zero-forcing is not needed for approximate optimality. Third, we derive an interesting trade-off between the receiver memory and the number of transmitters needed for approximately maximal performance. In particular, if each receiver can store a constant fraction of the content library, then only a constant number of transmitters are needed. Our solution to the caching problem requires formulating and solving a new communication problem, the symmetric multiple multicast X-channel, for which we provide an exact DoF characterization.",
"Over the past decade, the bulk of wireless traffic has shifted from speech to content. This shift creates the opportunity to cache part of the content in memories closer to the end users, for example in base stations. Most of the prior literature focuses on the reduction of load in the backhaul and core networks due to caching, i.e., on the benefits caching offers for the wireline communication link between the origin server and the caches. In this paper, we are instead interested in the benefits caching can offer for the wireless communication link between the caches and the end users. To quantify the gains of caching for this wireless link, we consider an interference channel in which each transmitter is equipped with an isolated cache memory. Communication takes place in two phases, a content placement phase followed by a content delivery phase. The objective is to design both the placement and the delivery phases to maximize the rate in the delivery phase in response to any possible user demands. Focusing on the three-user case, we show that through careful joint design of these phases, we can reap three distinct benefits from caching: a load balancing gain, an interference cancellation gain, and an interference alignment gain. In our proposed scheme, load balancing is achieved through a specific file splitting and placement, creating a particular pattern of content overlap at the caches. This overlap allows to implement interference cancellation. Further, it allows us to construct several virtual transmitters, each responsible for a part of the requested content, which increases interference alignment possibilities.",
""
]
} |
1904.05448 | 2936120942 | While human and group analysis have become an important area in last decades, some current and relevant applications involve to estimate future motion of pedestrians in real video sequences. This paper presents a method to provide motion estimation of real pedestrians in next seconds, using crowd simulation. Our method is based on Physics and heuristics and use BioCrowds as crowd simulation methodology to estimate future positions of people in video sequences. Results show that our method for estimation works well even for complex videos where events can happen. The maximum achieved average error is @math cm when estimating the future motion of 32 pedestrians with more than 2 seconds in advance. This paper discusses this and other results. | Some works have focused on the prediction of future behavior, as in @cite_16 , where the authors estimate the time taken by people to achieve their goals. For this, some characteristics need to be identified, such as regions of interest, origin and destination, people blocking the flow, as well as distribution of global density and speeds. Afterwards, using statistical calculations, the authors estimate the time to reach the goal. As said before, Khomchuk and collaborators @cite_10 use radar signatures (MD) through a supervised regression to estimate the mapping between the directions of motion and the corresponding MD signatures. Keller and Gavrila @cite_7 present a study on pedestrian path prediction at short subsecond time intervals. Their approaches are based on Gaussian process dynamical models and probabilistic hierarchical trajectory matching using a Kalman filter. Crowd simulation is the process by which the movement of many agents are calculated for the purpose of, for example, filling and animating virtual scenes @cite_4 @cite_17 or verifying the security of real environments @cite_8 . | {
"cite_N": [
"@cite_4",
"@cite_7",
"@cite_8",
"@cite_16",
"@cite_10",
"@cite_17"
],
"mid": [
"2605263008",
"2004641798",
"",
"2205807761",
"2501169257",
""
],
"abstract": [
"In current games, entire cities can be rendered in real time into massive virtual worlds. In addition to the enormous details of geometry, rendering, effects (e.g., particles), sound effects, and so on, nonplayable characters must also be animated and rendered, and they must interact with the environment and among themselves. Indeed, the computation time of all such data is expensive. Consequently, game designers should define priorities so that more resources can be allocated to generate better graphics, setting aside behavioral aspects. In huge environments, some of the actions behaviors that should be processed can be nonvisible to the players (occluded) or even visible but far away. Normally, in such cases, the common decision is to turn off such processing. However, hidden enemy behaviors that are not processed can result in nonrealistic feedback to the player. In this article, we aim to provide a method to preserve the motion of nonvisible characters while maintaining a compromise with the needed computational time of background behaviors. We apply this idea specifically in crowd collision behavior, proposing nonavoiding collision crowds. Such crowds do not have collision avoidance behaviors but preserve their motion as typical crowds. We propose a mathematical technique to describe how people are affected by others, so collision avoidance methods are not necessarily computed (they can be turned off, which leads to a reduction in the required computational time). Results show that our method replicates the behavior well (velocities, densities, and time) when compared to a free-of-collision method.",
"Future vehicle systems for active pedestrian safety will not only require a high recognition performance but also an accurate analysis of the developing traffic situation. In this paper, we present a study on pedestrian path prediction and action classification at short subsecond time intervals. We consider four representative approaches: two novel approaches (based on Gaussian process dynamical models and probabilistic hierarchical trajectory matching) that use augmented features derived from dense optical flow and two approaches as baseline that use positional information only (a Kalman filter and its extension to interacting multiple models). In experiments using stereo vision data obtained from a vehicle, we investigate the accuracy of path prediction and action classification at various time horizons, the effect of various errors (image localization, vehicle egomotion estimation), and the benefit of the proposed approaches. The scenario of interest is that of a crossing pedestrian, who might stop or continue walking at the road curbside. Results indicate similar performance of the four approaches on walking motion, with near-linear dynamics. During stopping, however, the two newly proposed approaches, with nonlinear and or higher order models and augmented motion features, achieve a more accurate position prediction of 10-50 cm at a time horizon of 0-0.77 s around the stopping event.",
"",
"In this paper, we target on the problem of estimating the statistic of pedestrian travel time within a period from an entrance to a destination in a crowded scene. Such estimation is based on the global distributions of crowd densities and velocities instead of complete trajectories of pedestrians, which cannot be obtained in crowded scenes. The proposed method is motivated by our statistical investigation into the correlations between travel time and global properties of crowded scenes. Active regions are created for each source-destination pair to model the probable walking regions over the corresponding source-destination traffic flow. Two sets of scene features are specially designed for modeling moving and stationary persons inside the active regions and their influences on pedestrian travel time. The estimation of pedestrian travel time provides valuable information for both crowd scene understanding and pedestrian behavior analysis, but was not sufficiently studied in literature. The effectiveness of the proposed pedestrian travel time estimation model is demonstrated through several surveillance applications, including dynamic scene monitoring, localization of regions blocking traffics, and detection of abnormal pedestrian behaviors. Many more valuable applications based on our method are to be explored in the future.",
"Micro-Doppler (MD)-based target classification capabilities of automotive radars can provide high reliability and short latency to future active safety automotive features. A large number of pedestrians surrounding a vehicle in practical urban scenarios mandate prioritization of their treatment level. Distinguishing between relevant pedestrians that are crossing the street or are within the vehicle path and those that are on the sidewalks and are moving along the vehicle route can significantly minimize the number of vehicle-to-pedestrian accidents. This work proposes a novel technique for estimating a pedestrian direction of motion that treats pedestrians as complex distributed targets and utilizes their MD radar signatures. The MD signatures are shown to be indicative of pedestrian direction of motion, and a supervised regression is used to estimate the mapping between the directions of motion and the corresponding MD signatures. In order to achieve higher regression performance, a state-of-the-art sparse dictionary learning-based feature extraction algorithm was adopted from the field of computer vision by drawing a parallel between the Doppler effect and the video temporal gradient. The performance of the proposed approach is evaluated in practical automotive scenario simulations, where a walking pedestrian is observed by a multiple-input multiple-output automotive radar with a two-dimensional rectangular array. The simulated data was generated using the statistical Boulic-Thalman human locomotion model. Accurate direction of motion estimation was achieved by using support vector regression and multilayer perceptron-based regression algorithms. The results show that the direction estimation error is less than 10° in 95 of the tested cases for pedestrian at a range of 100 m from the radar.",
""
]
} |
1904.05456 | 2252170720 | The era of big data has led to the emergence of new systems for real-time distributed stream processing, e.g., Apache Storm is one of the most popular stream processing systems in industry today. However, Storm, like many other stream processing systems lacks an intelligent scheduling mechanism. The default round-robin scheduling currently deployed in Storm disregards resource demands and availability, and can therefore be inefficient at times. We present R-Storm (Resource-Aware Storm), a system that implements resource-aware scheduling within Storm. R-Storm is designed to increase overall throughput by maximizing resource utilization while minimizing network latency. When scheduling tasks, R-Storm can satisfy both soft and hard resource constraints as well as minimizing network distance between components that communicate with each other. We evaluate R-Storm on set of micro-benchmark Storm applications as well as Storm applications used in production at Yahoo! Inc. From our experimental results we conclude that R-Storm achieves 30-47 higher throughput and 69-350 better CPU utilization than default Storm for the micro-benchmarks. For the Yahoo! Storm applications, R-Storm outperforms default Storm by around 50 based on overall throughput. We also demonstrate that R-Storm performs much better when scheduling multiple Storm applications than default Storm. | Not much work has been done in the area of resource-aware scheduling in Storm or distributed data stream systems. Some research work has been done in the space of scheduling for Hadoop MapReduce which share many similarities with real-time processing distributed system. In the work from Jorda @cite_24 , a resource-aware adaptive scheduling for a MapReduce Cluster is proposed and implemented. Their work is built upon the observation that there can be different workload jobs, and multiple users running at the same time on a cluster. By taking into account the memory and CPU capacities for each Task Tracker, the algorithm is able to find a job placement that maximizes a utility function while satisfying resource limits constraints. The algorithm is derived from a heuristic optimization algorithm to which is NP-hard. However, their work fails to take network into resource constraints which is a major bottleneck for streaming systems such as Storm. | {
"cite_N": [
"@cite_24"
],
"mid": [
"185167974"
],
"abstract": [
"We present a resource-aware scheduling technique for MapReduce multi-job workloads that aims at improving resource utilization across machines while observing completion time goals. Existing MapReduce schedulers define a static number of slots to represent the capacity of a cluster, creating a fixed number of execution slots per machine. This abstraction works for homogeneous workloads, but fails to capture the different resource requirements of individual jobs in multi-user environments. Our technique leverages job profiling information to dynamically adjust the number of slots on each machine, as well as workload placement across them, to maximize the resource utilization of the cluster. In addition, our technique is guided by user-provided completion time goals for each job. Source code of our prototype is available at [1]."
]
} |
1904.05059 | 2937946995 | Age estimation is a classic learning problem in computer vision. Many larger and deeper CNNs have been proposed with promising performance, such as AlexNet, VggNet, GoogLeNet and ResNet. However, these models are not practical for the embedded mobile devices. Recently, MobileNets and ShuffleNets have been proposed to reduce the number of parameters, yielding lightweight models. However, their representation has been weakened because of the adoption of depth-wise separable convolution. In this work, we investigate the limits of compact model for small-scale image and propose an extremely Compact yet efficient Cascade Context-based Age Estimation model(C3AE). This model possesses only 1 9 and 1 2000 parameters compared with MobileNets ShuffleNets and VggNet, while achieves competitive performance. In particular, we re-define age estimation problem by two-points representation, which is implemented by a cascade model. Moreover, to fully utilize the facial context information, multi-branch CNN network is proposed to aggregate multi-scale context. Experiments are carried out on three age estimation datasets. The state-of-the-art performance on compact model has been achieved with a relatively large margin. | As the increasing requirement of mobile embedded devices running deep learning, various efficient models such as GoogLeNet @cite_7 , SqueezeNet @cite_29 , ResNet @cite_3 and SENet @cite_33 , are designed to cater this wave. Recently, depth-wise convolution was adopted by MobileNets @cite_16 @cite_10 and ShuffleNets @cite_40 @cite_12 to reduce computation costs and model sizes. They were built primarily from depth-wise separable convolutions initially introduced in @cite_34 and subsequently used in Inception models @cite_35 @cite_14 to reduce the computation in the first few layers. In particular, the separation of filtering - applying convolution at each channel separately and combination - recombine the output of individual channels achieved fewer computations. MobileNet-V1 @cite_16 based on the depth-wise separable convolution explored some important design guidelines for an efficient model. ShuffleNet-V1 @cite_40 utilized novel point-wise group convolution and channel shuffle to reduce computation cost while maintaining accuracy. MobileNet-V2 @cite_21 proposed a novel inverted residual with linear bottleneck. ShuffleNet-V2 @cite_9 mainly analyzed the runtime performance of the model and give four guidelines for efficient network design. | {
"cite_N": [
"@cite_35",
"@cite_14",
"@cite_33",
"@cite_7",
"@cite_29",
"@cite_21",
"@cite_9",
"@cite_3",
"@cite_34",
"@cite_40",
"@cite_16",
"@cite_10",
"@cite_12"
],
"mid": [
"2183341477",
"",
"",
"2097117768",
"2279098554",
"2963163009",
"2883780447",
"2194775991",
"",
"2724359148",
"2612445135",
"",
""
],
"abstract": [
"Convolutional networks are at the core of most state of-the-art computer vision solutions for a wide variety of tasks. Since 2014 very deep convolutional networks started to become mainstream, yielding substantial gains in various benchmarks. Although increased model size and computational cost tend to translate to immediate quality gains for most tasks (as long as enough labeled data is provided for training), computational efficiency and low parameter count are still enabling factors for various use cases such as mobile vision and big-data scenarios. Here we are exploring ways to scale up networks in ways that aim at utilizing the added computation as efficiently as possible by suitably factorized convolutions and aggressive regularization. We benchmark our methods on the ILSVRC 2012 classification challenge validation set demonstrate substantial gains over the state of the art: 21:2 top-1 and 5:6 top-5 error for single frame evaluation using a network with a computational cost of 5 billion multiply-adds per inference and with using less than 25 million parameters. With an ensemble of 4 models and multi-crop evaluation, we report 3:5 top-5 error and 17:3 top-1 error on the validation set and 3:6 top-5 error on the official test set.",
"",
"",
"We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.",
"Recent research on deep neural networks has focused primarily on improving accuracy. For a given accuracy level, it is typically possible to identify multiple DNN architectures that achieve that accuracy level. With equivalent accuracy, smaller DNN architectures offer at least three advantages: (1) Smaller DNNs require less communication across servers during distributed training. (2) Smaller DNNs require less bandwidth to export a new model from the cloud to an autonomous car. (3) Smaller DNNs are more feasible to deploy on FPGAs and other hardware with limited memory. To provide all of these advantages, we propose a small DNN architecture called SqueezeNet. SqueezeNet achieves AlexNet-level accuracy on ImageNet with 50x fewer parameters. Additionally, with model compression techniques we are able to compress SqueezeNet to less than 0.5MB (510x smaller than AlexNet).",
"In this paper we describe a new mobile architecture, MobileNetV2, that improves the state of the art performance of mobile models on multiple tasks and benchmarks as well as across a spectrum of different model sizes. We also describe efficient ways of applying these mobile models to object detection in a novel framework we call SSDLite. Additionally, we demonstrate how to build mobile semantic segmentation models through a reduced form of DeepLabv3 which we call Mobile DeepLabv3. is based on an inverted residual structure where the shortcut connections are between the thin bottleneck layers. The intermediate expansion layer uses lightweight depthwise convolutions to filter features as a source of non-linearity. Additionally, we find that it is important to remove non-linearities in the narrow layers in order to maintain representational power. We demonstrate that this improves performance and provide an intuition that led to this design. Finally, our approach allows decoupling of the input output domains from the expressiveness of the transformation, which provides a convenient framework for further analysis. We measure our performance on ImageNet [1] classification, COCO object detection [2], VOC image segmentation [3]. We evaluate the trade-offs between accuracy, and number of operations measured by multiply-adds (MAdd), as well as actual latency, and the number of parameters.",
"Currently, the neural network architecture design is mostly guided by the indirect metric of computation complexity, i.e., FLOPs. However, the direct metric, e.g., speed, also depends on the other factors such as memory access cost and platform characterics. Thus, this work proposes to evaluate the direct metric on the target platform, beyond only considering FLOPs. Based on a series of controlled experiments, this work derives several practical guidelines for efficient network design. Accordingly, a new architecture is presented, called ShuffleNet V2. Comprehensive ablation experiments verify that our model is the state-of-the-art in terms of speed and accuracy tradeoff.",
"Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57 error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28 relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.",
"",
"We introduce an extremely computation-efficient CNN architecture named ShuffleNet, which is designed specially for mobile devices with very limited computing power (e.g., 10-150 MFLOPs). The new architecture utilizes two new operations, pointwise group convolution and channel shuffle, to greatly reduce computation cost while maintaining accuracy. Experiments on ImageNet classification and MS COCO object detection demonstrate the superior performance of ShuffleNet over other structures, e.g. lower top-1 error (absolute 7.8 ) than recent MobileNet on ImageNet classification task, under the computation budget of 40 MFLOPs. On an ARM-based mobile device, ShuffleNet achieves 13x actual speedup over AlexNet while maintaining comparable accuracy.",
"We present a class of efficient models called MobileNets for mobile and embedded vision applications. MobileNets are based on a streamlined architecture that uses depth-wise separable convolutions to build light weight deep neural networks. We introduce two simple global hyper-parameters that efficiently trade off between latency and accuracy. These hyper-parameters allow the model builder to choose the right sized model for their application based on the constraints of the problem. We present extensive experiments on resource and accuracy tradeoffs and show strong performance compared to other popular models on ImageNet classification. We then demonstrate the effectiveness of MobileNets across a wide range of applications and use cases including object detection, finegrain classification, face attributes and large scale geo-localization.",
"",
""
]
} |
1904.05019 | 2939052861 | Despite the fact that Second Order Similarity (SOS) has been used with significant success in tasks such as graph matching and clustering, it has not been exploited for learning local descriptors. In this work, we explore the potential of SOS in the field of descriptor learning by building upon the intuition that a positive pair of matching points should exhibit similar distances with respect to other points in the embedding space. Thus, we propose a novel regularization term, named Second Order Similarity Regularization (SOSR), that follows this principle. By incorporating SOSR into training, our learned descriptor achieves state-of-the-art performance on several challenging benchmarks containing distinct tasks ranging from local patch retrieval to structure from motion. Furthermore, by designing a von Mises-Fischer distribution based evaluation method, we link the utilization of the descriptor space to the matching performance, thus demonstrating the effectiveness of our proposed SOSR. Extensive experimental results, empirical evidence, and in-depth analysis are provided, indicating that SOSR can significantly boost the matching performance of the learned descriptor. | Early works on local patch description focused on low level processes such as gradient filters and intensity comparisons, including SIFT @cite_2 , GLOH @cite_29 , DAISY @cite_22 , DSP-SIFT @cite_15 and LIOP @cite_34 . A comprehensive review can be found in @cite_38 . | {
"cite_N": [
"@cite_38",
"@cite_22",
"@cite_29",
"@cite_2",
"@cite_15",
"@cite_34"
],
"mid": [
"2177274842",
"2169561155",
"",
"",
"1901447884",
"2163096995"
],
"abstract": [
"In this paper, we compare the performance of descriptors computed for local interest regions, as, for example, extracted by the Harris-Affine detector [Mikolajczyk, K and Schmid, C, 2004]. Many different descriptors have been proposed in the literature. It is unclear which descriptors are more appropriate and how their performance depends on the interest region detector. The descriptors should be distinctive and at the same time robust to changes in viewing conditions as well as to errors of the detector. Our evaluation uses as criterion recall with respect to precision and is carried out for different image transformations. We compare shape context [Belongie, S, et al, April 2002], steerable filters [Freeman, W and Adelson, E, Setp. 1991], PCA-SIFT [Ke, Y and Sukthankar, R, 2004], differential invariants [Koenderink, J and van Doorn, A, 1987], spin images [Lazebnik, S, et al, 2003], SIFT [Lowe, D. G., 1999], complex filters [Schaffalitzky, F and Zisserman, A, 2002], moment invariants [Van Gool, L, et al, 1996], and cross-correlation for different types of interest regions. We also propose an extension of the SIFT descriptor and show that it outperforms the original method. Furthermore, we observe that the ranking of the descriptors is mostly independent of the interest region detector and that the SIFT-based descriptors perform best. Moments and steerable filters show the best performance among the low dimensional descriptors.",
"Local image descriptors that are highly discriminative, computational efficient, and with low storage footprint have long been a dream goal of computer vision research. In this paper, we focus on learning such descriptors, which make use of the DAISY configuration and are simple to compute both sparsely and densely. We develop a new training set of match non-match image patches which improves on previous work. We test a wide variety of gradient and steerable filter based configurations and optimize over all parameters to obtain low matching errors for the descriptors. We further explore robust normalization, dimension reduction and dynamic range reduction to increase the discriminative power and yet reduce the storage requirement of the learned descriptors. All these enable us to obtain highly efficient local descriptors: e.g, 13.2 error at 13 bytes storage per descriptor, compared with 26.1 error at 128 bytes for SIFT.",
"",
"",
"We introduce a simple modification of local image descriptors, such as SIFT, based on pooling gradient orientations across different domain sizes, in addition to spatial locations. The resulting descriptor, which we call DSP-SIFT, outperforms other methods in wide-baseline matching benchmarks, including those based on convolutional neural networks, despite having the same dimension of SIFT and requiring no training.",
"This paper presents a novel method for feature description based on intensity order. Specifically, a Local Intensity Order Pattern(LIOP) is proposed to encode the local ordinal information of each pixel and the overall ordinal information is used to divide the local patch into subregions which are used for accumulating the LIOPs respectively. Therefore, both local and overall intensity ordinal information of the local patch are captured by the proposed LIOP descriptor so as to make it a highly discriminative descriptor. It is shown that the proposed descriptor is not only invariant to monotonic intensity changes and image rotation but also robust to many other geometric and photometric transformations such as viewpoint change, image blur and JEPG compression. The proposed descriptor has been evaluated on the standard Oxford dataset and four additional image pairs with complex illumination changes. The experimental results show that the proposed descriptor obtains a significant improvement over the existing state-of-the-art descriptors."
]
} |
1904.05019 | 2939052861 | Despite the fact that Second Order Similarity (SOS) has been used with significant success in tasks such as graph matching and clustering, it has not been exploited for learning local descriptors. In this work, we explore the potential of SOS in the field of descriptor learning by building upon the intuition that a positive pair of matching points should exhibit similar distances with respect to other points in the embedding space. Thus, we propose a novel regularization term, named Second Order Similarity Regularization (SOSR), that follows this principle. By incorporating SOSR into training, our learned descriptor achieves state-of-the-art performance on several challenging benchmarks containing distinct tasks ranging from local patch retrieval to structure from motion. Furthermore, by designing a von Mises-Fischer distribution based evaluation method, we link the utilization of the descriptor space to the matching performance, thus demonstrating the effectiveness of our proposed SOSR. Extensive experimental results, empirical evidence, and in-depth analysis are provided, indicating that SOSR can significantly boost the matching performance of the learned descriptor. | With the emergence of annotated patch datasets @cite_4 , a significant amount of data-driven methods focused on improving hand-crafted representations using machine learning methods. The authors of @cite_11 @cite_33 use linear projections to learn discriminative descriptors, while convex optimization is used for learning an optimal descriptor sampling configuration in @cite_24 . BinBoost @cite_14 is trained based on a boosting framework, and in RFD @cite_16 , the most discriminative receptive fields are learned based on the labeled training data. BOLD @cite_27 uses patch specific adaptive online selection of binary intensity tests. | {
"cite_N": [
"@cite_14",
"@cite_4",
"@cite_33",
"@cite_24",
"@cite_27",
"@cite_16",
"@cite_11"
],
"mid": [
"2126338221",
"",
"1992159177",
"2007178811",
"1906968832",
"2097934584",
"2089888558"
],
"abstract": [
"Binary key point descriptors provide an efficient alternative to their floating-point competitors as they enable faster processing while requiring less memory. In this paper, we propose a novel framework to learn an extremely compact binary descriptor we call Bin Boost that is very robust to illumination and viewpoint changes. Each bit of our descriptor is computed with a boosted binary hash function, and we show how to efficiently optimize the different hash functions so that they complement each other, which is key to compactness and robustness. The hash functions rely on weak learners that are applied directly to the image patches, which frees us from any intermediate representation and lets us automatically learn the image gradient pooling configuration of the final descriptor. Our resulting descriptor significantly outperforms the state-of-the-art binary descriptors and performs similarly to the best floating-point descriptors at a fraction of the matching time and memory footprint.",
"",
"In this paper, we present Linear Discriminant Projections (LDP) for reducing dimensionality and improving discriminability of local image descriptors. We place LDP into the context of state-of-the-art discriminant projections and analyze its properties. LDP requires a large set of training data with point-to-point correspondence ground truth. We demonstrate that training data produced by a simulation of image transformations leads to nearly the same results as the real data with correspondence ground truth. This makes it possible to apply LDP as well as other discriminant projection approaches to the problems where the correspondence ground truth is not available, such as image categorization. We perform an extensive experimental evaluation on standard data sets in the context of image matching and categorization. We demonstrate that LDP enables significant dimensionality reduction of local descriptors and performance increases in different applications. The results improve upon the state-of-the-art recognition performance with simultaneous dimensionality reduction from 128 to 30.",
"The objective of this work is to learn descriptors suitable for the sparse feature detectors used in viewpoint invariant matching. We make a number of novel contributions towards this goal. First, it is shown that learning the pooling regions for the descriptor can be formulated as a convex optimisation problem selecting the regions using sparsity. Second, it is shown that descriptor dimensionality reduction can also be formulated as a convex optimisation problem, using Mahalanobis matrix nuclear norm regularisation. Both formulations are based on discriminative large margin learning constraints. As the third contribution, we evaluate the performance of the compressed descriptors, obtained from the learnt real-valued descriptors by binarisation. Finally, we propose an extension of our learning formulations to a weakly supervised case, which allows us to learn the descriptors from unannotated image collections. It is demonstrated that the new learning methods improve over the state of the art in descriptor learning on the annotated local patches data set of and unannotated photo collections of .",
"In this paper we propose a novel approach to generate a binary descriptor optimized for each image patch independently. The approach is inspired by the linear discriminant embedding that simultaneously increases inter and decreases intra class distances. A set of discriminative and uncorrelated binary tests is established from all possible tests in an offline training process. The patch adapted descriptors are then efficiently built online from a subset of tests which lead to lower intra class distances thus a more robust descriptor. A patch descriptor consists of two binary strings where one represents the results of the tests and the other indicates the subset of the patch-related robust tests that are used for calculating a masked Hamming distance. Our experiments on three different benchmarks demonstrate improvements in matching performance, and illustrate that per-patch optimization outperforms global optimization.",
"Feature description for local image patch is widely used in computer vision. While the conventional way to design local descriptor is based on expert experience and knowledge, learning-based methods for designing local descriptor become more and more popular because of their good performance and data-driven property. This paper proposes a novel data-driven method for designing binary feature descriptor, which we call receptive fields descriptor (RFD). Technically, RFD is constructed by thresholding responses of a set of receptive fields, which are selected from a large number of candidates according to their distinctiveness and correlations in a greedy way. Using two different kinds of receptive fields (namely rectangular pooling area and Gaussian pooling area) for selection, we obtain two binary descriptors @math and @math .accordingly. Image matching experiments on the well-known patch data set and Oxford data set demonstrate that RFD significantly outperforms the state-of-the-art binary descriptors, and is comparable with the best float-valued descriptors at a fraction of processing time. Finally, experiments on object recognition tasks confirm that both @math and @math successfully bridge the performance gap between binary descriptors and their floating-point competitors.",
"In this paper, we explore methods for learning local image descriptors from training data. We describe a set of building blocks for constructing descriptors which can be combined together and jointly optimized so as to minimize the error of a nearest-neighbor classifier. We consider both linear and nonlinear transforms with dimensionality reduction, and make use of discriminant learning techniques such as Linear Discriminant Analysis (LDA) and Powell minimization to solve for the parameters. Using these techniques, we obtain descriptors that exceed state-of-the-art performance with low dimensionality. In addition to new experiments and recommendations for descriptor learning, we are also making available a new and realistic ground truth data set based on multiview stereo data."
]
} |
1904.05019 | 2939052861 | Despite the fact that Second Order Similarity (SOS) has been used with significant success in tasks such as graph matching and clustering, it has not been exploited for learning local descriptors. In this work, we explore the potential of SOS in the field of descriptor learning by building upon the intuition that a positive pair of matching points should exhibit similar distances with respect to other points in the embedding space. Thus, we propose a novel regularization term, named Second Order Similarity Regularization (SOSR), that follows this principle. By incorporating SOSR into training, our learned descriptor achieves state-of-the-art performance on several challenging benchmarks containing distinct tasks ranging from local patch retrieval to structure from motion. Furthermore, by designing a von Mises-Fischer distribution based evaluation method, we link the utilization of the descriptor space to the matching performance, thus demonstrating the effectiveness of our proposed SOSR. Extensive experimental results, empirical evidence, and in-depth analysis are provided, indicating that SOSR can significantly boost the matching performance of the learned descriptor. | While the recent improvements in the field of learning CNN patch descriptors are significant, methods mentioned above are limited to optimizing the measured by @math distances of positive and negative pairs, and the potential of has not been exploited in the area of descriptor learning. On the other hand, graph matching algorithms @cite_30 @cite_1 @cite_7 have been developed based on due to its robustness to shape distortions. Furthermore, @cite_5 proves that using can achieve better cluster performance. Thus, our key idea is to introduce constraints at the training stage for robust patch description. | {
"cite_N": [
"@cite_30",
"@cite_5",
"@cite_1",
"@cite_7"
],
"mid": [
"1587878450",
"",
"2103375950",
"2076080556"
],
"abstract": [
"Graph matching is an essential problem in computer vision and machine learning. In this paper, we introduce a random walk view on the problem and propose a robust graph matching algorithm against outliers and deformation. Matching between two graphs is formulated as node selection on an association graph whose nodes represent candidate correspondences between the two graphs. The solution is obtained by simulating random walks with reweighting jumps enforcing the matching constraints on the association graph. Our algorithm achieves noise-robust graph matching by iteratively updating and exploiting the confidences of candidate correspondences. In a practical sense, our work is of particular importance since the real-world matching problem is made difficult by the presence of noise and outliers. Extensive and comparative experiments demonstrate that it outperforms the state-of-the-art graph matching algorithms especially in the presence of outliers and deformation.",
"",
"Graph matching is widely used in a variety of scientific fields, including computer vision, due to its powerful performance, robustness, and generality. Its computational complexity, however, limits the permissible size of input graphs in practice. Therefore, in real-world applications, the initial construction of graphs to match becomes a critical factor for the matching performance, and often leads to unsatisfactory results. In this paper, to resolve the issue, we propose a novel progressive framework which combines probabilistic progression of graphs with matching of graphs. The algorithm efficiently re-estimates in a Bayesian manner the most plausible target graphs based on the current matching result, and guarantees to boost the matching objective at the subsequent graph matching. Experimental evaluation demonstrates that our approach effectively handles the limits of conventional graph matching and achieves significant improvement in challenging image matching problems.",
"Graph matching plays a central role in solving correspondence problems in computer vision. Graph matching problems that incorporate pair-wise constraints can be cast as a quadratic assignment problem (QAP). Unfortunately, QAP is NP-hard and many algorithms have been proposed to solve different relaxations. This paper presents factorized graph matching (FGM), a novel framework for interpreting and optimizing graph matching problems. In this work we show that the affinity matrix can be factorized as a Kronecker product of smaller matrices. There are three main benefits of using this factorization in graph matching: (1) There is no need to compute the costly (in space and time) pair-wise affinity matrix; (2) The factorization provides a taxonomy for graph matching and reveals the connection among several methods; (3) Using the factorization we derive a new approximation of the original problem that improves state-of-the-art algorithms in graph matching. Experimental results in synthetic and real databases illustrate the benefits of FGM. The code is available at http: humansensing.cs.cmu.edu fgm."
]
} |
1904.05161 | 2938827034 | A growing set of applications consider the process of network formation by using subgraphs as a tool for generating the network topology. One of the pressing research challenges is thus to be able to use these subgraphs to understand the network topology of information cascades which ultimately paves the way to theorize about how information spreads over time. In this paper, we make the first attempt at using network motifs to understand whether or not they can be used as generative elements for the diffusion network organization during different phases of the cascade lifecycle. In doing so, we propose a motif percolation-based algorithm that uses network motifs to measure the extent to which they can represent the temporal cascade network organization. We compare two phases of the cascade lifecycle from the perspective of diffusion-- the phase of steep growth and the phase of inhibition prior to its saturation. Our experiments on a set of cascades from the Weibo platform and with 5-node motifs demonstrate that there are only a few specific motif patterns with triads that are able to characterize the spreading process and hence the network organization during the inhibition region better than during the phase of high growth. In contrast, we do not find compelling results for the phase of steep growth. | * -3mm Our work is closely related to the field of community detection which enjoys a prolific literature. In particular, community detection in static networks has received much attention from multiple disciplines including sociology and computer science. This has resulted in a diverse set of existing methods including traditional network structure-based studies @cite_13 , optimization-based techniques @cite_14 @cite_10 @cite_2 , label propagation methods @cite_1 @cite_17 and propinquity @cite_4 . Also, some other studies have studied the evolution of communities over time and in dynamic networks. The work of @cite_9 for example, detected natural communities that were stable to small perturbations of the input data. Later, new communities detected were matched to earlier snapshots using the natural community structure tree. Another work of @cite_6 which in essence is related to our work, proposed a method to detect communities in dynamic networks based on the -Clique Percolation Method (CPM). In their approach, communities were defined as adjacent k-cliques that share k-1 nodes. More specifically, two nodes belong to the same community if they can be connected through adjacent k-cliques. * -4mm | {
"cite_N": [
"@cite_13",
"@cite_14",
"@cite_4",
"@cite_9",
"@cite_1",
"@cite_6",
"@cite_2",
"@cite_10",
"@cite_17"
],
"mid": [
"1971421925",
"1970301364",
"2119625792",
"2129043771",
"2132202037",
"19833001",
"2119222198",
"2098166511",
"2118059508"
],
"abstract": [
"A number of recent studies have focused on the statistical properties of networked systems such as social networks and the Worldwide Web. Researchers have concentrated particularly on a few properties that seem to be common to many networks: the small-world property, power-law degree distributions, and network transitivity. In this article, we highlight another property that is found in many networks, the property of community structure, in which network nodes are joined together in tightly knit groups, between which there are only looser connections. We propose a method for detecting such communities, built around the idea of using centrality indices to find community boundaries. We test our method on computer-generated and real-world graphs whose community structure is already known and find that the method detects this known structure with high sensitivity and reliability. We also apply the method to two networks whose community structure is not well known—a collaboration network and a food web—and find that it detects significant and informative community divisions in both cases.",
"Community structure is one of the main structural features of networks, revealing both their internal organization and the similarity of their elementary units. Despite the large variety of methods proposed to detect communities in graphs, there is a big need for multi-purpose techniques, able to handle different types of datasets and the subtleties of community structure. In this paper we present OSLOM (Order Statistics Local Optimization Method), the first method capable to detect clusters in networks accounting for edge directions, edge weights, overlapping communities, hierarchies and community dynamics. It is based on the local optimization of a fitness function expressing the statistical significance of clusters with respect to random fluctuations, which is estimated with tools of Extreme and Order Statistics. OSLOM can be used alone or as a refinement procedure of partitions covers delivered by other techniques. We have also implemented sequential algorithms combining OSLOM with other fast techniques, so that the community structure of very large networks can be uncovered. Our method has a comparable performance as the best existing algorithms on artificial benchmark graphs. Several applications on real networks are shown as well. OSLOM is implemented in a freely available software (http: www.oslom.org), and we believe it will be a valuable tool in the analysis of networks.",
"Graphs or networks can be used to model complex systems. Detecting community structures from large network data is a classic and challenging task. In this paper, we propose a novel community detection algorithm, which utilizes a dynamic process by contradicting the network topology and the topology-based propinquity, where the propinquity is a measure of the probability for a pair of nodes involved in a coherent community structure. Through several rounds of mutual reinforcement between topology and propinquity, the community structures are expected to naturally emerge. The overlapping vertices shared between communities can also be easily identified by an additional simple postprocessing. To achieve better efficiency, the propinquity is incrementally calculated. We implement the algorithm on a vertex-oriented bulk synchronous parallel(BSP) model so that the mining load can be distributed on thousands of machines. We obtained interesting experimental results on several real network data.",
"We are interested in tracking changes in large-scale data by periodically creating an agglomerative clustering and examining the evolution of clusters (communities) over time. We examine a large real-world data set: the NEC CiteSeer database, a linked network of >250,000 papers. Tracking changes over time requires a clustering algorithm that produces clusters stable under small perturbations of the input data. However, small perturbations of the CiteSeer data lead to significant changes to most of the clusters. One reason for this is that the order in which papers within communities are combined is somewhat arbitrary. However, certain subsets of papers, called natural communities, correspond to real structure in the CiteSeer database and thus appear in any clustering. By identifying the subset of clusters that remain stable under multiple clustering runs, we get the set of natural communities that we can track over time. We demonstrate that such natural communities allow us to identify emerging communities and track temporal changes in the underlying structure of our network data.",
"Community detection and analysis is an important methodology for understanding the organization of various real-world networks and has applications in problems as diverse as consensus formation in social communities or the identification of functional modules in biochemical networks. Currently used algorithms that identify the community structures in large-scale real-world networks require a priori information such as the number and sizes of communities or are computationally expensive. In this paper we investigate a simple label propagation algorithm that uses the network structure alone as its guide and requires neither optimization of a predefined objective function nor prior information about the communities. In our algorithm every node is initialized with a unique label and at every step each node adopts the label that most of its neighbors currently have. In this iterative process densely connected groups of nodes form a consensus on a unique label to form communities. We validate the algorithm by applying it to networks whose community structures are known. We also demonstrate that the algorithm takes an almost linear time and hence it is computationally less expensive than what was possible so far.",
"The rich set of interactions between individuals in the society results in complex community structure, capturing highly connected circles of friends, fam- ilies, or professional cliques in a social network. Due to the frequent changes in the activity and communication patterns of individuals, the associated social and communication network is subject to constant evolution. The cohesive groups of people in such networks can grow by recruiting new members, or contract by loos- ing members; two (or more) groups may merge into a single community, while a large enough social group can split into several smaller ones; new communities are born and old ones may disappear. We discuss a new algorithm based on a clique percolation technique, that allows to investigate in detail the time dependence of communities on a large scale and as such, to uncover basic relationships of the sta- tistical features of community evolution. According to the results, the behaviour of smaller collaborative or friendship circles and larger communities, e.g., institutions show significant differences. Social groups containing only a few members persist longer on average when the fluctuations of the members is small. In contrast, we find that the condition for stability for large communities is continuous changes in their membership, allowing for the possibility that after some time practically all members are exchanged.",
"In this paper, we introduce a game-theoretic framework to address the community detection problem based on the structures of social networks. We formulate the dynamics of community formation as a strategic game called community formation game: Given an underlying social graph, we assume that each node is a selfish agent who selects communities to join or leave based on her own utility measurement. A community structure can be interpreted as an equilibrium of this game. We formulate the agents' utility by the combination of a gain function and a loss function. We allow each agent to select multiple communities, which naturally captures the concept of \"overlapping communities\". We propose a gain function based on the modularity concept introduced by Newman (Proc Natl Acad Sci 103(23):8577---8582, 2006), and a simple loss function that reflects the intrinsic costs incurred when people join the communities. We conduct extensive experiments under this framework, and our results show that our algorithm is effective in identifying overlapping communities, and are often better then other algorithms we evaluated especially when many people belong to multiple communities. To the best of our knowledge, this is the first time the community detection problem is addressed by a game-theoretic framework that considers community formation as the result of individual agents' rational behaviors.",
"Most complex networks demonstrate a significant property 'community structure', meaning that the network nodes are often joined together in tightly knit groups or communities, while there are only looser connections between them. Detecting these groups is of great importance and has immediate applications, especially in the popular online social networks like Facebook and Twitter. Many of these networks are divided into overlapping communities, i.e. communities with nodes belonging to more than one community simultaneously. Unfortunately most of the works cannot detect such communities. In this paper, we consider the formation of communities in social networks as an iterative game in a multiagent environment, in which, each node is regarded as an agent trying to be in the communities with members structurally equivalent to her. Remarkable results on the real world and benchmark graphs show efficiency of our approach in detecting overlapping communities compared to the other similar methods.",
"Studies of community structure and evolution in large social networks require a fast and accurate algorithm for community detection. As the size of analyzed communities grows, complexity of the community detection algorithm needs to be kept close to linear. The Label Propagation Algorithm (LPA) has the benefits of nearly-linear running time and easy implementation, thus it forms a good basis for efficient community detection methods. In this paper, we propose new update rule and label propagation criterion in LPA to improve both its computational efficiency and the quality of communities that it detects. The speed is optimized by avoiding unnecessary updates performed by the original algorithm. This change reduces significantly (by order of magnitude for large networks) the number of iterations that the algorithm executes. We also evaluate our generalization of the LPA update rule that takes into account, with varying strength, connections to the neighborhood of a node considering a new label. Experiments on computer generated networks and a wide range of social networks show that our new rule improves the quality of the detected communities compared to those found by the original LPA. The benefit of considering positive neighborhood strength is pronounced especially on real-world networks containing sufficiently large fraction of nodes with high clustering coefficient."
]
} |
1904.04975 | 2938145879 | Re-identifying a person across multiple disjoint camera views is important for intelligent video surveillance, smart retailing and many other applications. However, existing person re-identification (ReID) methods are challenged by the ubiquitous occlusion over persons and suffer from performance degradation. This paper proposes a novel occlusion-robust and alignment-free model for occluded person ReID and extends its application to realistic and crowded scenarios. The proposed model first leverages the full convolution network (FCN) and pyramid pooling to extract spatial pyramid features. Then an alignment-free matching approach, namely Foreground-aware Pyramid Reconstruction (FPR), is developed to accurately compute matching scores between occluded persons, despite their different scales and sizes. FPR uses the error from robust reconstruction over spatial pyramid features to measure similarities between two persons. More importantly, we design an occlusion-sensitive foreground probability generator that focuses more on clean human body parts to refine the similarity computation with less contamination from occlusion. The FPR is easily embedded into any end-to-end person ReID models. The effectiveness of the proposed method is clearly demonstrated by the experimental results (Rank-1 accuracy) on three occluded person datasets: Partial REID (78.30 ), Partial iLIDS (68.08 ) and Occluded REID (81.00 ); and three benchmark person datasets: Market1501 (95.42 ), DukeMTMC (88.64 ) and CUHK03 (76.08 ) | @cite_19 @cite_8 @cite_12 use person masks that contain body shape information to help remove the background clutters at pixel-level for person re-identification. For example, Kalayeh @cite_19 proposed a model that integrates human semantic parsing in person re-identification. It is similar to @cite_19 , Qi @cite_6 combined source images with person masks as the inputs to remove the appearance variations (illumination, pose, occlusion, etc.). @cite_15 @cite_2 @cite_16 utilize the skeleton as an external cue to effectively relieve the part misalignment problem by locating each part using person landmarks. For instance, Su @cite_18 proposed a Pose-driven Deep Convolutional (PDC) model to learn improved feature extractors and matching models in an end-to-end manner. The PDC can explicitly leverage the human part cues to alleviate the identification difficulties caused by pose variations. Suh @cite_28 proposed a two-stream network, which consists of an appearance map extraction stream and a body part map extraction stream. Following the two streams, a part-aligned feature map was obtained by a bilinear mapping of the corresponding local appearance and body part descriptors. Although these approaches can indeed address occlusion problem, they heavily depend on accurate pedestrian segmentation, and also cost much time to infer the external cues. | {
"cite_N": [
"@cite_18",
"@cite_8",
"@cite_28",
"@cite_6",
"@cite_19",
"@cite_2",
"@cite_15",
"@cite_16",
"@cite_12"
],
"mid": [
"2963438548",
"",
"2796664602",
"2797394071",
"2963805953",
"",
"2798429327",
"",
""
],
"abstract": [
"Feature extraction and matching are two crucial components in person Re-Identification (ReID). The large pose deformations and the complex view variations exhibited by the captured person images significantly increase the difficulty of learning and matching of the features from person images. To overcome these difficulties, in this work we propose a Pose-driven Deep Convolutional (PDC) model to learn improved feature extraction and matching models from end to end. Our deep architecture explicitly leverages the human part cues to alleviate the pose variations and learn robust feature representations from both the global image and different local parts. To match the features from global human body and local body parts, a pose driven feature weighting sub-network is further designed to learn adaptive feature fusions. Extensive experimental analyses and results on three popular datasets demonstrate significant performance improvements of our model over all published state-of-the-art methods.",
"",
"We propose a novel network that learns a part-aligned representation for person re-identification. It handles the body part misalignment problem, that is, body parts are misaligned across human detections due to pose viewpoint change and unreliable detection. Our model consists of a two-stream network (one stream for appearance map extraction and the other one for body part map extraction) and a bilinear-pooling layer that generates and spatially pools a part-aligned map. Each local feature of the part-aligned map is obtained by a bilinear mapping of the corresponding local appearance and body part descriptors. Our new representation leads to a robust image matching similarity, which is equivalent to an aggregation of the local similarities of the corresponding body parts combined with the weighted appearance similarity. This part-aligned representation reduces the part misalignment problem significantly. Our approach is also advantageous over other pose-guided representations (e.g., extracting representations over the bounding box of each body part) by learning part descriptors optimal for person re-identification. For training the network, our approach does not require any part annotation on the person re-identification dataset. Instead, we simply initialize the part sub-stream using a pre-trained sub-network of an existing pose estimation network, and train the whole network to minimize the re-identification loss. We validate the effectiveness of our approach by demonstrating its superiority over the state-of-the-art methods on the standard benchmark datasets, including Market-1501, CUHK03, CUHK01 and DukeMTMC, and standard video dataset MARS.",
"Person retrieval faces many challenges including cluttered background, appearance variations (e.g., illumination, pose, occlusion) among different camera views and the similarity among different person's images. To address these issues, we put forward a novel mask based deep ranking neural network with a skipped fusing layer. Firstly, to alleviate the problem of cluttered background, masked images with only the foreground regions are incorporated as input in the proposed neural network. Secondly, to reduce the impact of the appearance variations, the multi-layer fusion scheme is developed to obtain more discriminative fine-grained information. Lastly, considering person retrieval is a special image retrieval task, we propose a novel ranking loss to optimize the whole network. The proposed ranking loss can further mitigate the interference problem of similar negative samples when producing ranking results. The extensive experiments validate the superiority of the proposed method compared with the state-of-the-art methods on many benchmark datasets.",
"Person re-identification is a challenging task mainly due to factors such as background clutter, pose, illumination and camera point of view variations. These elements hinder the process of extracting robust and discriminative representations, hence preventing different identities from being successfully distinguished. To improve the representation learning, usually local features from human body parts are extracted. However, the common practice for such a process has been based on bounding box part detection. In this paper, we propose to adopt human semantic parsing which, due to its pixel-level accuracy and capability of modeling arbitrary contours, is naturally a better alternative. Our proposed SPReID integrates human semantic parsing in person re-identification and not only considerably outperforms its counter baseline, but achieves state-of-the-art performance. We also show that, by employing a simple yet effective training strategy, standard popular deep convolutional architectures such as Inception-V3 and ResNet-152, with no modification, while operating solely on full image, can dramatically outperform current state-of-the-art. Our proposed methods improve state-of-the-art person re-identification on: Market-1501 [48] by 17 in mAP and 6 in rank-1, CUHK03 [24] by 4 in rank-1 and DukeMTMC-reID [50] by 24 in mAP and 10 in rank-1.",
"",
"Person re-identification (ReID) is an important task in the field of intelligent security. A key challenge is how to capture human pose variations, while existing benchmarks (i.e., Market1501, DukeMTMC-reID, CUHK03, etc.) do NOT provide sufficient pose coverage to train a robust ReID system. To address this issue, we propose a pose-transferrable person ReID framework which utilizes posetransferred sample augmentations (i.e., with ID supervision) to enhance ReID model training. On one hand, novel training samples with rich pose variations are generated via transferring pose instances from MARS dataset, and they are added into the target dataset to facilitate robust training. On the other hand, in addition to the conventional discriminator of GAN (i.e., to distinguish between REAL FAKE samples), we propose a novel guider sub-network which encourages the generated sample (i.e., with novel pose) towards better satisfying the ReID loss (i.e., cross-entropy ReID loss, triplet ReID loss). In the meantime, an alternative optimization procedure is proposed to train the proposed Generator-Guider-Discriminator network. Experimental results on Market-1501, DukeMTMC-reID and CUHK03 show that our method achieves great performance improvement, and outperforms most state-of-the-art methods without elaborate designing the ReID model.",
"",
""
]
} |
1904.04989 | 2939497660 | Data association-based multiple object tracking (MOT) involves multiple separated modules processed or optimized differently, which results in complex method design and requires non-trivial tuning of parameters. In this paper, we present an end-to-end model, named FAMNet, where Feature extraction, Affinity estimation and Multi-dimensional assignment are refined in a single network. All layers in FAMNet are designed differentiable thus can be optimized jointly to learn the discriminative features and higher-order affinity model for robust MOT, which is supervised by the loss directly from the assignment ground truth. We also integrate single object tracking technique and a dedicated target management scheme into the FAMNet-based tracking system to further recover false negatives and inhibit noisy target candidates generated by the external detector. The proposed method is evaluated on a diverse set of benchmarks including MOT2015, MOT2017, KITTI-Car and UA-DETRAC, and achieves promising performance on all of them in comparison with state-of-the-arts. | Multiple object tracking (MOT) has been an active research area for decades, and many methods have been investigated for this topic. Recently, the most popular framework for MOT is the tracking-by-detection. Traditional methods primarily focus on solving the data association problem using such as Hungarian algorithm @cite_25 @cite_28 @cite_14 , network flow @cite_10 @cite_8 @cite_33 and multiple hypotheses tracking @cite_37 @cite_24 on various of affinity estimation schemes. Higher-order affinity provides the global and discriminative information that is not available in pairwise association. In order to utilize it, MOT is usually treated as the MDA problem. Collins @cite_11 proposes a block ICM-like method for the MDA to incorporate higher-order motion model. The method iteratively solves bipartite assignments alternatively while keeping other assignments fixed. @cite_26 , MDA is formulated as the rank-1 tensor approximation problem where a dedicated power iteration with unite @math normalization is proposed to find the optimal solution. Our work is closely related to the MDA formulation, especially @cite_26 . | {
"cite_N": [
"@cite_37",
"@cite_14",
"@cite_26",
"@cite_33",
"@cite_8",
"@cite_28",
"@cite_24",
"@cite_10",
"@cite_25",
"@cite_11"
],
"mid": [
"2739374836",
"2122469558",
"",
"2111644456",
"1528063097",
"2964019074",
"2237765446",
"1921895346",
"2252355370",
"2115734113"
],
"abstract": [
"Tracking-by-detection has become a popular tracking paradigm in recent years. Due to the fact that detections within this framework are regarded as points in the tracking process, it brings data association ambiguities, especially in crowded scenarios. To cope with this issue, we extended the multiple hypothesis tracking approach by incorporating a novel enhancing detection model that included detection-scene analysis and detection-detection analysis; the former models the scene by using dense confidential detections and handles false trajectories, while the latter estimates the correlations between individual detections and improves the ability to deal with close object hypotheses in crowded scenarios. Our approach was tested on the MOT16 benchmark and achieved competitive results with current state-of-the-art trackers.",
"We present a detection-based three-level hierarchical association approach to robustly track multiple objects in crowded environments from a single camera. At the low level, reliable tracklets (i.e. short tracks for further analysis) are generated by linking detection responses based on conservative affinity constraints. At the middle level, these tracklets are further associated to form longer tracklets based on more complex affinity measures. The association is formulated as a MAP problem and solved by the Hungarian algorithm. At the high level, entries, exits and scene occluders are estimated using the already computed tracklets, which are used to refine the final trajectories. This approach is applied to the pedestrian class and evaluated on two challenging datasets. The experimental results show a great improvement in performance compared to previous methods.",
"",
"We propose a network flow based optimization method for data association needed for multiple object tracking. The maximum-a-posteriori (MAP) data association problem is mapped into a cost-flow network with a non-overlap constraint on trajectories. The optimal data association is found by a min-cost flow algorithm in the network. The network is augmented to include an explicit occlusion model(EOM) to track with long-term inter-object occlusions. A solution to the EOM-based network is found by an iterative approach built upon the original algorithm. Initialization and termination of trajectories and potential false observations are modeled by the formulation intrinsically. The method is efficient and does not require hypotheses pruning. Performance is compared with previous results on two public pedestrian datasets to show its improvement.",
"Data association is an essential component of any human tracking system. The majority of current methods, such as bipartite matching, incorporate a limited-temporal-locality of the sequence into the data association problem, which makes them inherently prone to IDswitches and difficulties caused by long-term occlusion, cluttered background, and crowded scenes.We propose an approach to data association which incorporates both motion and appearance in a global manner. Unlike limited-temporal-locality methods which incorporate a few frames into the data association problem, we incorporate the whole temporal span and solve the data association problem for one object at a time, while implicitly incorporating the rest of the objects. In order to achieve this, we utilize Generalized Minimum Clique Graphs to solve the optimization problem of our data association method. Our proposed method yields a better formulated approach to data association which is supported by our superior results. Experiments show the proposed method makes significant improvements in tracking in the diverse sequences of Town Center [1], TUD-crossing [2], TUD-Stadtmitte [2], PETS2009 [3], and a new sequence called Parking Lot compared to the state of the art methods.",
"The main challenge of online multi-object tracking is to reliably associate object trajectories with detections in each video frame based on their tracking history. In this work, we propose the Recurrent Autoregressive Network (RAN), a temporal generative modeling framework to characterize the appearance and motion dynamics of multiple objects over time. The RAN couples an external memory and an internal memory. The external memory explicitly stores previous inputs of each trajectory in a time window, while the internal memory learns to summarize long-term tracking history and associate detections by processing the external memory. We conduct experiments on the MOT 2015 and 2016 datasets to demonstrate the robustness of our tracking method in highly crowded and occluded scenes. Our method achieves top-ranked results on the two benchmarks.",
"This paper revisits the classical multiple hypotheses tracking (MHT) algorithm in a tracking-by-detection framework. The success of MHT largely depends on the ability to maintain a small list of potential hypotheses, which can be facilitated with the accurate object detectors that are currently available. We demonstrate that a classical MHT implementation from the 90's can come surprisingly close to the performance of state-of-the-art methods on standard benchmark datasets. In order to further utilize the strength of MHT in exploiting higher-order information, we introduce a method for training online appearance models for each track hypothesis. We show that appearance models can be learned efficiently via a regularized least squares framework, requiring only a few extra operations for each hypothesis branch. We obtain state-of-the-art results on popular tracking-by-detection datasets such as PETS and the recent MOT challenge.",
"In this paper we show that multiple object tracking (MOT) can be formulated in a framework, where the detection and data-association are performed simultaneously. Our method allows us to overcome the confinements of data association based MOT approaches; where the performance is dependent on the object detection results provided at input level. At the core of our method lies structured learning which learns a model for each target and infers the best location of all targets simultaneously in a video clip. The inference of our structured learning is done through a new Target Identity-aware Network Flow (TINF), where each node in the network encodes the probability of each target identity belonging to that node. The proposed Lagrangian relaxation optimization finds the high quality solution to the network. During optimization a soft spatial constraint is enforced between the nodes of the graph which helps reducing the ambiguity caused by nearby targets with similar appearance in crowded scenarios. We show that automatically detecting and tracking targets in a single framework can help resolve the ambiguities due to frequent occlusion and heavy articulation of targets. Our experiments involve challenging yet distinct datasets and show that our method can achieve results better than the state-of-art.",
"This paper explores a pragmatic approach to multiple object tracking where the main focus is to associate objects efficiently for online and realtime applications. To this end, detection quality is identified as a key factor influencing tracking performance, where changing the detector can improve tracking by up to 18.9 . Despite only using a rudimentary combination of familiar techniques such as the Kalman Filter and Hungarian algorithm for the tracking components, this approach achieves an accuracy comparable to state-of-the-art online trackers. Furthermore, due to the simplicity of our tracking method, the tracker updates at a rate of 260 Hz which is over 20x faster than other state-of-the-art trackers.",
"We present an iterative approximate solution to the multidimensional assignment problem under general cost functions. The method maintains a feasible solution at every step, and is guaranteed to converge. It is similar to the iterated conditional modes (ICM) algorithm, but applied at each step to a block of variables representing correspondences between two adjacent frames, with the optimal conditional mode being calculated exactly as the solution to a two-frame linear assignment problem. Experiments with ground-truthed trajectory data show that the method outperforms both network-flow data association and greedy recursive filtering using a constant velocity motion model."
]
} |
1904.04989 | 2939497660 | Data association-based multiple object tracking (MOT) involves multiple separated modules processed or optimized differently, which results in complex method design and requires non-trivial tuning of parameters. In this paper, we present an end-to-end model, named FAMNet, where Feature extraction, Affinity estimation and Multi-dimensional assignment are refined in a single network. All layers in FAMNet are designed differentiable thus can be optimized jointly to learn the discriminative features and higher-order affinity model for robust MOT, which is supervised by the loss directly from the assignment ground truth. We also integrate single object tracking technique and a dedicated target management scheme into the FAMNet-based tracking system to further recover false negatives and inhibit noisy target candidates generated by the external detector. The proposed method is evaluated on a diverse set of benchmarks including MOT2015, MOT2017, KITTI-Car and UA-DETRAC, and achieves promising performance on all of them in comparison with state-of-the-arts. | Recently, deep learning is explored with increasing popularity in MOT with great success. Most recent solutions rely on it as a powerful discriminative technique @cite_39 @cite_9 @cite_49 @cite_36 . @cite_47 propose to use DNN based Re-ID techniques for affinity estimations. They include lift edges that connect two candidates spanning multiple frames to capture the long-term affinity. @cite_43 , recurrent neural networks (RNN) and long short-term memory (LSTM) is adapted to model the higher-order discriminative clue. Those methods learn the networks in a separate process with the manually fabricated affinity training samples. | {
"cite_N": [
"@cite_36",
"@cite_9",
"@cite_39",
"@cite_43",
"@cite_49",
"@cite_47"
],
"mid": [
"2895150009",
"2964015640",
"2604679602",
"2951063106",
"2749203358",
"2739491435"
],
"abstract": [
"In this paper, we propose an online Multi-Object Tracking (MOT) approach which integrates the merits of single object tracking and data association methods in a unified framework to handle noisy detections and frequent interactions between targets. Specifically, for applying single object tracking in MOT, we introduce a cost-sensitive tracking loss based on the state-of-the-art visual tracker, which encourages the model to focus on hard negative distractors during online learning. For data association, we propose Dual Matching Attention Networks (DMAN) with both spatial and temporal attention mechanisms. The spatial attention module generates dual attention maps which enable the network to focus on the matching patterns of the input image pair, while the temporal attention module adaptively allocates different levels of attention to different samples in the tracklet to suppress noisy observations. Experimental results on the MOT benchmark datasets show that the proposed algorithm performs favorably against both online and offline trackers in terms of identity-preserving metrics.",
"This paper introduces a novel approach to the task of data association within the context of pedestrian tracking, by introducing a two-stage learning scheme to match pairs of detections. First, a Siamese convolutional neural network (CNN) is trained to learn descriptors encoding local spatio-temporal structures between the two input image patches, aggregating pixel values and optical flow information. Second, a set of contextual features derived from the position and size of the compared input patches are combined with the CNN output by means of a gradient boosting classifier to generate the final matching probability. This learning approach is validated by using a linear programming based multi-person tracker showing that even a simple and efficient tracker may outperform much more complex models when fed with our learned matching probabilities. Results on publicly available sequences show that our method meets state-of-the-art standards in multiple people tracking.",
"Online multi-object tracking aims at estimating the tracks of multiple objects instantly with each incoming frame and the information provided up to the moment. It still remains a difficult problem in complex scenes, because of the large ambiguity in associating multiple objects in consecutive frames and the low discriminability between objects appearances. In this paper, we propose a robust online multi-object tracking method that can handle these difficulties effectively. We first define the tracklet confidence using the detectability and continuity of a tracklet, and decompose a multi-object tracking problem into small subproblems based on the tracklet confidence. We then solve the online multi-object tracking problem by associating tracklets and detections in different ways according to their confidence values. Based on this strategy, tracklets sequentially grow with online-provided detections, and fragmented tracklets are linked up with others without any iterative and expensive association steps. For more reliable association between tracklets and detections, we also propose a deep appearance learning method to learn a discriminative appearance model from large training datasets, since the conventional appearance learning methods do not provide rich representation that can distinguish multiple objects with large appearance variations. In addition, we combine online transfer learning for improving appearance discriminability by adapting the pre-trained deep model during online tracking. Experiments with challenging public datasets show distinct performance improvement over other state-of-the-arts batch and online tracking methods, and prove the effect and usefulness of the proposed methods for online multi-object tracking.",
"We present a multi-cue metric learning framework to tackle the popular yet unsolved Multi-Object Tracking (MOT) problem. One of the key challenges of tracking methods is to effectively compute a similarity score that models multiple cues from the past such as object appearance, motion, or even interactions. This is particularly challenging when objects get occluded or share similar appearance properties with surrounding objects. To address this challenge, we cast the problem as a metric learning task that jointly reasons on multiple cues across time. Our framework learns to encode long-term temporal dependencies across multiple cues with a hierarchical Recurrent Neural Network. We demonstrate the strength of our approach by tracking multiple objects using their appearance, motion, and interactions. Our method outperforms previous works by a large margin on multiple publicly available datasets including the challenging MOT benchmark.",
"We propose Quadruplet Convolutional Neural Networks (Quad-CNN) for multi-object tracking, which learn to associate object detections across frames using quadruplet losses. The proposed networks consider target appearances together with their temporal adjacencies for data association. Unlike conventional ranking losses, the quadruplet loss enforces an additional constraint that makes temporally adjacent detections more closely located than the ones with large temporal gaps. We also employ a multi-task loss to jointly learn object association and bounding box regression for better localization. The whole network is trained end-to-end. For tracking, the target association is performed by minimax label propagation using the metric learned from the proposed network. We evaluate performance of our multi-object tracking algorithm on public MOT Challenge datasets, and achieve outstanding results.",
"Tracking multiple persons in a monocular video of a crowded scene is a challenging task. Humans can master it even if they loose track of a person locally by re-identifying the same person based on their appearance. Care must be taken across long distances, as similar-looking persons need not be identical. In this work, we propose a novel graph-based formulation that links and clusters person hypotheses over time by solving an instance of a minimum cost lifted multicut problem. Our model generalizes previous works by introducing a mechanism for adding long-range attractive connections between nodes in the graph without modifying the original set of feasible solutions. This allows us to reward tracks that assign detections of similar appearance to the same person in a way that does not introduce implausible solutions. To effectively match hypotheses over longer temporal gaps we develop new deep architectures for re-identification of people. They combine holistic representations extracted with deep networks and body pose layout obtained with a state-of-the-art pose estimation model. We demonstrate the effectiveness of our formulation by reporting a new state-of-the-art for the MOT16 benchmark. The code and pre-trained models are publicly available."
]
} |
1904.04989 | 2939497660 | Data association-based multiple object tracking (MOT) involves multiple separated modules processed or optimized differently, which results in complex method design and requires non-trivial tuning of parameters. In this paper, we present an end-to-end model, named FAMNet, where Feature extraction, Affinity estimation and Multi-dimensional assignment are refined in a single network. All layers in FAMNet are designed differentiable thus can be optimized jointly to learn the discriminative features and higher-order affinity model for robust MOT, which is supervised by the loss directly from the assignment ground truth. We also integrate single object tracking technique and a dedicated target management scheme into the FAMNet-based tracking system to further recover false negatives and inhibit noisy target candidates generated by the external detector. The proposed method is evaluated on a diverse set of benchmarks including MOT2015, MOT2017, KITTI-Car and UA-DETRAC, and achieves promising performance on all of them in comparison with state-of-the-arts. | Some recent works have gone further to tentatively solve MOT in an entirely end-to-end fashion. Ondruska and Posner @cite_40 introduce the RNN for the task to estimate the candidate state. Although this work is demonstrated on the synthetic sensor data and no explicit data association is applied, it firstly shows the efficacy of using RNN for an end-to-end solution. @cite_6 propose an RNN-LSTM based online framework to integrate both motion affinity estimation and bipartite association into the deep learning network. They use LSTM to solve the data association target by target at each frame where the constrains in data association are not explicitly built into the network but learned from training data. For both works, only the occupancy status of targets are considered, the informative appearance clue is not utilized. Different from their methods, we propose an MDA sub-network which handles both the data association and the constrains, and our affinity fuses both the appearance and motion clue for better discriminability. | {
"cite_N": [
"@cite_40",
"@cite_6"
],
"mid": [
"2963121817",
"2963063317"
],
"abstract": [
"This paper presents to the best of our knowledge the first end-to-end object tracking approach which directly maps from raw sensor input to object tracks in sensor space without requiring any feature engineering or system identification in the form of plant or sensor models. Specifically, our system accepts a stream of raw sensor data at one end and, in real-time, produces an estimate of the entire environment state at the output including even occluded objects. We achieve this by framing the problem as a deep learning task and exploit sequence models in the form of recurrent neural networks to learn a mapping from sensor measurements to object tracks. In particular, we propose a learning method based on a form of input dropout which allows learning in an unsupervised manner, only based on raw, occluded sensor data without access to ground-truth annotations. We demonstrate our approach using a synthetic dataset designed to mimic the task of tracking objects in 2D laser data – as commonly encountered in robotics applications – and show that it learns to track many dynamic objects despite occlusions and the presence of sensor noise.",
"We present a novel approach to online multi-target tracking based on recurrent neural networks (RNNs). Tracking multiple objects in real-world scenes involves many challenges, including a) an a-priori unknown and time-varying number of targets, b) a continuous state estimation of all present targets, and c) a discrete combinatorial problem of data association. Most previous methods involve complex models that require tedious tuning of parameters. Here, we propose for the first time, an end-to-end learning approach for online multi-target tracking. Existing deep learning methods are not designed for the above challenges and cannot be trivially applied to the task. Our solution addresses all of the above points in a principled way. Experiments on both synthetic and real data show promising results obtained at 300 Hz on a standard CPU, and pave the way towards future research in this direction."
]
} |
1904.04978 | 2940087192 | Automatic Check-Out (ACO) receives increased interests in recent years. An important component of the ACO system is the visual item counting, which recognize the categories and counts of the items chosen by the customers. However, the training of such a system is challenged by the domain adaptation problem, in which the training data are images from isolated items while the testing images are for collections of items. Existing methods solve this problem with data augmentation using synthesized images, but the image synthesis leads to unreal images that affect the training process. In this paper, we propose a new data priming method to solve the domain adaptation problem. Specifically, we first use pre-augmentation data priming, in which we remove distracting background from the training images and select images with realistic view angles by the pose pruning method. In the post-augmentation step, we train a data priming network using detection and counting collaborative learning, and select more reliable images from testing data to train the final visual item tallying network. Experiments on the large scale Retail Product Checkout (RPC) dataset demonstrate the superiority of the proposed method, i.e., we achieve 80.51 checkout accuracy compared with 56.68 of the baseline methods. | Salient object detection @cite_25 @cite_9 @cite_10 @cite_2 @cite_23 is to segment the main object in the image for pre-processing. Li al @cite_25 obtain the saliency map based on the multi-scale features extracted from CNN models. Hu al @cite_10 propose a saliency detection method based on the compactness hypothesis that assumes salient regions are more compact than background from the perspectives of both color layout and texture layout. Liu al @cite_9 develop a two-stage deep network, where a coarse prediction map is produced and followed by a recurrent CNN to refine the details of the prediction map hierarchically and progressively. Tang and Wu @cite_2 develop multiple single-scale fully convolutional networks integrated chained connections to generate saliency prediction results from coarse to fine. Recently, Hou al @cite_23 take full advantage of multi-level and multi-scale features extracted from fully CNNs, and introduce short connections to the skip-layer structures within the holistically-nested edge detector. | {
"cite_N": [
"@cite_9",
"@cite_23",
"@cite_2",
"@cite_10",
"@cite_25"
],
"mid": [
"2461475918",
"2569272946",
"2766915143",
"2486803733",
"1894057436"
],
"abstract": [
"Traditional1 salient object detection models often use hand-crafted features to formulate contrast and various prior knowledge, and then combine them artificially. In this work, we propose a novel end-to-end deep hierarchical saliency network (DHSNet) based on convolutional neural networks for detecting salient objects. DHSNet first makes a coarse global prediction by automatically learning various global structured saliency cues, including global contrast, objectness, compactness, and their optimal combination. Then a novel hierarchical recurrent convolutional neural network (HRCNN) is adopted to further hierarchically and progressively refine the details of saliency maps step by step via integrating local context information. The whole architecture works in a global to local and coarse to fine manner. DHSNet is directly trained using whole images and corresponding ground truth saliency masks. When testing, saliency maps can be generated by directly and efficiently feedforwarding testing images through the network, without relying on any other techniques. Evaluations on four benchmark datasets and comparisons with other 11 state-of-the-art algorithms demonstrate that DHSNet not only shows its significant superiority in terms of performance, but also achieves a real-time speed of 23 FPS on modern GPUs.",
"Recent progress on salient object detection is substantial, benefiting mostly from the explosive development of Convolutional Neural Networks (CNNs). Semantic segmentation and salient object detection algorithms developed lately have been mostly based on Fully Convolutional Neural Networks (FCNs). There is still a large room for improvement over the generic FCN models that do not explicitly deal with the scale-space problem. The Holistically-Nested Edge Detector (HED) provides a skip-layer structure with deep supervision for edge and boundary detection, but the performance gain of HED on saliency detection is not obvious. In this paper, we propose a new salient object detection method by introducing short connections to the skip-layer structures within the HED architecture. Our framework takes full advantage of multi-level and multi-scale features extracted from FCNs, providing more advanced representations at each layer, a property that is critically needed to perform segment detection. Our method produces state-of-the-art results on 5 widely tested salient object detection benchmarks, with advantages in terms of efficiency (0.08 seconds per image), effectiveness, and simplicity over the existing algorithms. Beyond that, we conduct an exhaustive analysis of the role of training data on performance. We provide a training set for future research and fair comparisons.",
"In this paper, we proposed a novel method for effective salient object detection by designing a chained multi-scale fully convolutional network (CMSFCN). CMSFCN contained multiple single-scale fully convolutional networks (SSFCNs), which were integrated successively by using chained connections and generated saliency prediction results from coarse to fine. The chained connections not only combined the saliency prediction result from previous SSFCN with the input image of current SSFCN, but also combined the intermediate features from previous SSFCN and current SSFCN. With these chained connections, the sequential SSFCNs in CMSFCN automatically learned complemental and discriminative features to improve the saliency predictions progressively. Therefore, after jointly training CMSFCN with an end-to-end manner, precise saliency prediction results were produced under a coarse-to-fine behaviour. Compared with seven state-of-the-art CNN based salient object detection approaches over five benchmark datasets, experimental results demonstrated the efficiency and effectiveness of CMSFCN.",
"In recent years, the object-level saliency detection has attracted much research attention, due to its usefulness in many high-level tasks. Existing methods are mostly based on the contrast hypothesis, which regards the regions with high contrast in a certain context as salient objects. Although the contrast hypothesis is effective in many scenarios, it cannot handle some difficult cases. As a remedy to address the weakness of contrast hypothesis, we propose a novel compactness hypothesis, which assumes salient regions are more compact than background from the perspectives of both color layout and texture layout. Based on the compactness hypotheses, we implement an effective object-level saliency detection method. In the proposed method, we first construct a weak saliency map based on the compact hypotheses, then collect samples from the weak saliency map to train a dedicated classifier. This classifier is applied on each individual pixel of the input image to produce a confidence score. Finally, the confidence scores are used to form a saliency map. This process is carried out at different scales, and the corresponding results are integrated into the formation of the final saliency map. The proposed approach is evaluated on eight benchmark data sets, where it delivers the competitive performance compared with the state-of-the-art methods.",
"Visual saliency is a fundamental problem in both cognitive and computational sciences, including computer vision. In this paper, we discover that a high-quality visual saliency model can be learned from multiscale features extracted using deep convolutional neural networks (CNNs), which have had many successes in visual recognition tasks. For learning such saliency models, we introduce a neural network architecture, which has fully connected layers on top of CNNs responsible for feature extraction at three different scales. We then propose a refinement method to enhance the spatial coherence of our saliency results. Finally, aggregating multiple saliency maps computed for different levels of image segmentation can further boost the performance, yielding saliency maps better than those generated from a single segmentation. To promote further research and evaluation of visual saliency models, we also construct a new large database of 4447 challenging images and their pixelwise saliency annotations. Experimental results demonstrate that our proposed method is capable of achieving state-of-the-art performance on all public benchmarks, improving the F-Measure by 5.0 and 13.2 respectively on the MSRA-B dataset and our new dataset (HKU-IS), and lowering the mean absolute error by 5.7 and 35.1 respectively on these two datasets."
]
} |
1904.04978 | 2940087192 | Automatic Check-Out (ACO) receives increased interests in recent years. An important component of the ACO system is the visual item counting, which recognize the categories and counts of the items chosen by the customers. However, the training of such a system is challenged by the domain adaptation problem, in which the training data are images from isolated items while the testing images are for collections of items. Existing methods solve this problem with data augmentation using synthesized images, but the image synthesis leads to unreal images that affect the training process. In this paper, we propose a new data priming method to solve the domain adaptation problem. Specifically, we first use pre-augmentation data priming, in which we remove distracting background from the training images and select images with realistic view angles by the pose pruning method. In the post-augmentation step, we train a data priming network using detection and counting collaborative learning, and select more reliable images from testing data to train the final visual item tallying network. Experiments on the large scale Retail Product Checkout (RPC) dataset demonstrate the superiority of the proposed method, i.e., we achieve 80.51 checkout accuracy compared with 56.68 of the baseline methods. | Data augmentation is a common method used in deep network training to deal with training data shortage. Recently, generative models including variational auto-encoder (VAE) @cite_18 @cite_24 and generative adversarial networks (GANs) @cite_31 @cite_11 are used to synthesize images similar to those in realistic scenes for data augmentation. Oord al @cite_18 propose a new conditional image generation method based on the Pixel-CNN structure. It can be conditioned feature vectors obtained from descriptive labels or tags, or latent embeddings created by other networks. @cite_24 , a layered VAE model with disentangled latent variables is proposed to generate images from visual attributes. Besides, different from VAE, Goodfellow al @cite_31 estimate generative models via an adversarial process of two models, where the generative model captures the data distribution, and the discriminative model estimates the probability that a sample came from the training data rather than the generative model. Recently, the CycleGAN model @cite_11 is to learn the mapping between an input image and an output image in different styles. | {
"cite_N": [
"@cite_24",
"@cite_18",
"@cite_31",
"@cite_11"
],
"mid": [
"2963567641",
"2963636093",
"2099471712",
"2962793481"
],
"abstract": [
"This paper investigates a novel problem of generating images from visual attributes. We model the image as a composite of foreground and background and develop a layered generative model with disentangled latent variables that can be learned end-to-end using a variational auto-encoder. We experiment with natural images of faces and birds and demonstrate that the proposed models are capable of generating realistic and diverse samples with disentangled latent representations. We use a general energy minimization algorithm for posterior inference of latent variables given novel images. Therefore, the learned generative models show excellent quantitative and visual results in the tasks of attribute-conditioned image reconstruction and completion.",
"This work explores conditional image generation with a new image density model based on the PixelCNN architecture. The model can be conditioned on any vector, including descriptive labels or tags, or latent embeddings created by other networks. When conditioned on class labels from the ImageNet database, the model is able to generate diverse, realistic scenes representing distinct animals, objects, landscapes and structures. When conditioned on an embedding produced by a convolutional network given a single image of an unseen face, it generates a variety of new portraits of the same person with different facial expressions, poses and lighting conditions. We also show that conditional PixelCNN can serve as a powerful decoder in an image autoencoder. Additionally, the gated convolutional layers in the proposed model improve the log-likelihood of PixelCNN to match the state-of-the-art performance of PixelRNN on ImageNet, with greatly reduced computational cost.",
"We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a mistake. This framework corresponds to a minimax two-player game. In the space of arbitrary functions G and D, a unique solution exists, with G recovering the training data distribution and D equal to ½ everywhere. In the case where G and D are defined by multilayer perceptrons, the entire system can be trained with backpropagation. There is no need for any Markov chains or unrolled approximate inference networks during either training or generation of samples. Experiments demonstrate the potential of the framework through qualitative and quantitative evaluation of the generated samples.",
"Image-to-image translation is a class of vision and graphics problems where the goal is to learn the mapping between an input image and an output image using a training set of aligned image pairs. However, for many tasks, paired training data will not be available. We present an approach for learning to translate an image from a source domain X to a target domain Y in the absence of paired examples. Our goal is to learn a mapping G : X → Y such that the distribution of images from G(X) is indistinguishable from the distribution Y using an adversarial loss. Because this mapping is highly under-constrained, we couple it with an inverse mapping F : Y → X and introduce a cycle consistency loss to push F(G(X)) ≈ X (and vice versa). Qualitative results are presented on several tasks where paired training data does not exist, including collection style transfer, object transfiguration, season transfer, photo enhancement, etc. Quantitative comparisons against several prior methods demonstrate the superiority of our approach."
]
} |
1904.04978 | 2940087192 | Automatic Check-Out (ACO) receives increased interests in recent years. An important component of the ACO system is the visual item counting, which recognize the categories and counts of the items chosen by the customers. However, the training of such a system is challenged by the domain adaptation problem, in which the training data are images from isolated items while the testing images are for collections of items. Existing methods solve this problem with data augmentation using synthesized images, but the image synthesis leads to unreal images that affect the training process. In this paper, we propose a new data priming method to solve the domain adaptation problem. Specifically, we first use pre-augmentation data priming, in which we remove distracting background from the training images and select images with realistic view angles by the pose pruning method. In the post-augmentation step, we train a data priming network using detection and counting collaborative learning, and select more reliable images from testing data to train the final visual item tallying network. Experiments on the large scale Retail Product Checkout (RPC) dataset demonstrate the superiority of the proposed method, i.e., we achieve 80.51 checkout accuracy compared with 56.68 of the baseline methods. | In training deep learning models, due to many factors, there exists a shift between the domains of the training and testing data that can degrade the performance. Domain adaptation uses labeled data in one source domains to apply to testing data in a target domain. Recently there have been several domain adaptation methods for visual data. @cite_22 , the authors learn deep features such that they are not only discriminative for the main learning task on the source domain but invariant with respect to the shift between the domains. Saito al @cite_16 propose an asymmetric tri-training method for unsupervised domain adaptation, where unlabeled samples are assigned to pseudo-labels and train neural networks as if they are true labels. @cite_27 , a novel Manifold Embedded Distribution Alignment method is proposed to learn the domain-invariant classifier with the principle of structural risk minimization while performing dynamic distribution alignment. The work of @cite_5 adapts the Faster R-CNN @cite_6 with both image and instance level domain adaptation components to reduce the domain discrepancy. Qi al @cite_12 propose a covariant multimodal attention based multimodal domain adaptation method by adaptively fusing attended features of different modalities. | {
"cite_N": [
"@cite_22",
"@cite_6",
"@cite_27",
"@cite_5",
"@cite_16",
"@cite_12"
],
"mid": [
"2963826681",
"2613718673",
"2884771968",
"2964115968",
"2594718649",
"2897482938"
],
"abstract": [
"Top-performing deep architectures are trained on massive amounts of labeled data. In the absence of labeled data for a certain task, domain adaptation often provides an attractive option given that labeled data of similar nature but from a different domain (e.g. synthetic images) are available. Here, we propose a new approach to domain adaptation in deep architectures that can be trained on large amount of labeled data from the source domain and large amount of unlabeled data from the target domain (no labeled target-domain data is necessary). As the training progresses, the approach promotes the emergence of \"deep\" features that are (i) discriminative for the main learning task on the source domain and (ii) invariant with respect to the shift between the domains. We show that this adaptation behaviour can be achieved in almost any feed-forward model by augmenting it with few standard layers and a simple new gradient reversal layer. The resulting augmented architecture can be trained using standard back propagation. Overall, the approach can be implemented with little effort using any of the deep-learning packages. The method performs very well in a series of image classification experiments, achieving adaptation effect in the presence of big domain shifts and outperforming previous state-of-the-art on Office datasets.",
"State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet [7] and Fast R-CNN [5] have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully-convolutional network that simultaneously predicts object bounds and objectness scores at each position. RPNs are trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. With a simple alternating optimization, RPN and Fast R-CNN can be trained to share convolutional features. For the very deep VGG-16 model [19], our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007 (73.2 mAP) and 2012 (70.4 mAP) using 300 proposals per image. Code is available at https: github.com ShaoqingRen faster_rcnn.",
"Visual domain adaptation aims to learn robust classifiers for the target domain by leveraging knowledge from a source domain. Existing methods either attempt to align the cross-domain distributions, or perform manifold subspace learning. However, there are two significant challenges: (1) degenerated feature transformation, which means that distribution alignment is often performed in the original feature space, where feature distortions are hard to overcome. On the other hand, subspace learning is not sufficient to reduce the distribution divergence. (2) unevaluated distribution alignment, which means that existing distribution alignment methods only align the marginal and conditional distributions with equal importance, while they fail to evaluate the different importance of these two distributions in real applications. In this paper, we propose a Manifold Embedded Distribution Alignment (MEDA) approach to address these challenges. MEDA learns a domain-invariant classifier in Grassmann manifold with structural risk minimization, while performing dynamic distribution alignment to quantitatively account for the relative importance of marginal and conditional distributions. To the best of our knowledge, MEDA is the first attempt to perform dynamic distribution alignment for manifold domain adaptation. Extensive experiments demonstrate that MEDA shows significant improvements in classification accuracy compared to state-of-the-art traditional and deep methods.",
"Object detection typically assumes that training and test data are drawn from an identical distribution, which, however, does not always hold in practice. Such a distribution mismatch will lead to a significant performance drop. In this work, we aim to improve the cross-domain robustness of object detection. We tackle the domain shift on two levels: 1) the image-level shift, such as image style, illumination, etc., and 2) the instance-level shift, such as object appearance, size, etc. We build our approach based on the recent state-of-the-art Faster R-CNN model, and design two domain adaptation components, on image level and instance level, to reduce the domain discrepancy. The two domain adaptation components are based on @math -divergence theory, and are implemented by learning a domain classifier in adversarial training manner. The domain classifiers on different levels are further reinforced with a consistency regularization to learn a domain-invariant region proposal network (RPN) in the Faster R-CNN model. We evaluate our newly proposed approach using multiple datasets including Cityscapes, KITTI, SIM10K, etc. The results demonstrate the effectiveness of our proposed approach for robust object detection in various domain shift scenarios.",
"It is important to apply models trained on a large number of labeled samples to different domains because collecting many labeled samples in various domains is expensive. To learn discriminative representations for the target domain, we assume that artificially labeling the target samples can result in a good representation. Tri-training leverages three classifiers equally to provide pseudo-labels to unlabeled samples; however, the method does not assume labeling samples generated from a different domain. In this paper, we propose the use of an asymmetric tri-training method for unsupervised domain adaptation, where we assign pseudo-labels to unlabeled samples and train the neural networks as if they are true labels. In our work, we use three networks asymmetrically, and by asymmetric, we mean that two networks are used to label unlabeled target samples, and one network is trained by the pseudo-labeled samples to obtain target-discriminative representations. Our proposed method was shown to achieve a state-of-the-art performance on the benchmark digit recognition datasets for domain adaptation.",
"Domain adaptation aims to train a model on labeled data from a source domain while minimizing test error on a target domain. Most of existing domain adaptation methods only focus on reducing domain shift of single-modal data. In this paper, we consider a new problem of multimodal domain adaptation and propose a unified framework to solve it. The proposed multimodal domain adaptation neural networks(MDANN) consist of three important modules. (1) A covariant multimodal attention is designed to learn a common feature representation for multiple modalities. (2) A fusion module adaptively fuses attended features of different modalities. (3) Hybrid domain constraints are proposed to comprehensively learn domain-invariant features by constraining single modal features, fused features, and attention scores. Through jointly attending and fusing under an adversarial objective, the most discriminative and domain-adaptive parts of the features are adaptively fused together. Extensive experimental results on two real-world cross-domain applications (emotion recognition and cross-media retrieval) demonstrate the effectiveness of the proposed method."
]
} |
1904.04978 | 2940087192 | Automatic Check-Out (ACO) receives increased interests in recent years. An important component of the ACO system is the visual item counting, which recognize the categories and counts of the items chosen by the customers. However, the training of such a system is challenged by the domain adaptation problem, in which the training data are images from isolated items while the testing images are for collections of items. Existing methods solve this problem with data augmentation using synthesized images, but the image synthesis leads to unreal images that affect the training process. In this paper, we propose a new data priming method to solve the domain adaptation problem. Specifically, we first use pre-augmentation data priming, in which we remove distracting background from the training images and select images with realistic view angles by the pose pruning method. In the post-augmentation step, we train a data priming network using detection and counting collaborative learning, and select more reliable images from testing data to train the final visual item tallying network. Experiments on the large scale Retail Product Checkout (RPC) dataset demonstrate the superiority of the proposed method, i.e., we achieve 80.51 checkout accuracy compared with 56.68 of the baseline methods. | To date, there only exists a handful of related datasets for grocery product classification @cite_21 , recognition @cite_4 @cite_8 @cite_3 @cite_7 , segmentation @cite_0 and tallying @cite_29 . | {
"cite_N": [
"@cite_4",
"@cite_7",
"@cite_8",
"@cite_29",
"@cite_21",
"@cite_3",
"@cite_0"
],
"mid": [
"2140571654",
"2555440590",
"2127095579",
"2914924014",
"2166828762",
"",
"2963539956"
],
"abstract": [
"In this paper, a new image set, called the Surrey Object Image Library (SOIL-47) is introduced, on which the performance of two colour-based object recognition methods is evaluated. The data was collected specifically for testing colour-based recognition algorithm and is publicly available. In the conducted experiments on SOIL-47, we evaluate two recognition algoritms; the Multimodal Neighbourhood Signature (MNS) approach and a method based on a Attributed Relational Graph (ARG). The MNS approach represents object appearance by measurements computed from image neighbourhoods with a multimodal colour density function. The ARG approach computes a graph of affine invariant measurements of the colour and shape of segmented image regions. Using only a single model image of each of the 47 objects, MNS performed well even for extreme test views close to degrees. The ARG method assumes a locally planar surface, therefore a second experiment was conducted on a subset of box-like objects of SOIL-47. MNS performance was fairly stable, outperforming ARG for most viewing angles.. Note, that this is the fir st systematic test of MNS with controlled 3D viewpoint change.",
"With the increasing performance of machine learning techniques in the last few years, the computer vision and robotics communities have created a large number of datasets for benchmarking object recognition tasks. These datasets cover a large spectrum of natural images and object categories, making them not only useful as a testbed for comparing machine learning approaches, but also a great resource for bootstrapping different domain-specific perception and robotic systems. One such domain is domestic environments, where an autonomous robot has to recognize a large variety of everyday objects such as groceries. This is a challenging task due to the large variety of objects and products, and where there is great need for real-world training data that goes beyond product images available online. In this paper, we address this issue and present a dataset consisting of 5,000 images covering 25 different classes of groceries, with at least 97 images per class. We collected all images from real-world settings at different stores and apartments. In contrast to existing groceries datasets, our dataset includes a large variety of perspectives, lighting conditions, and degrees of clutter. Overall, our images contain thousands of different object instances. It is our hope that machine learning and robotics researchers find this dataset of use for training, testing, and bootstrapping their approaches. As a baseline classifier to facilitate comparison, we re-trained the CaffeNet architecture (an adaptation of the well-known AlexNet) on our dataset and achieved a mean accuracy of 78.9 . We release this trained model along with the code and data splits we used in our experiments.",
"The problem of using pictures of objects captured under ideal imaging conditions (here referred to as in vitro) to recognize objects in natural environments (in situ) is an emerging area of interest in computer vision and pattern recognition. Examples of tasks in this vein include assistive vision systems for the blind and object recognition for mobile robots; the proliferation of image databases on the web is bound to lead to more examples in the near future. Despite its importance, there is still a need for a freely available database to facilitate study of this kind of training testing dichotomy. In this work one of our contributions is a new multimedia database of 120 grocery products, GroZi-120. For every product, two different recordings are available: in vitro images extracted from the web, and in situ images extracted from camcorder video collected inside a grocery store. As an additional contribution, we present the results of applying three commonly used object recognition detection algorithms (color histogram matching, SIFT matching, and boosted Haar-like features) to the dataset. Finally, we analyze the successes and failures of these algorithms against product type and imaging conditions, both in terms of recognition rate and localization accuracy, in order to suggest ways forward for further research in this domain.",
"Over recent years, emerging interest has occurred in integrating computer vision technology into the retail industry. Automatic checkout (ACO) is one of the critical problems in this area which aims to automatically generate the shopping list from the images of the products to purchase. The main challenge of this problem comes from the large scale and the fine-grained nature of the product categories as well as the difficulty for collecting training images that reflect the realistic checkout scenarios due to continuous update of the products. Despite its significant practical and research value, this problem is not extensively studied in the computer vision community, largely due to the lack of a high-quality dataset. To fill this gap, in this work we propose a new dataset to facilitate relevant research. Our dataset enjoys the following characteristics: (1) It is by far the largest dataset in terms of both product image quantity and product categories. (2) It includes single-product images taken in a controlled environment and multi-product images taken by the checkout system. (3) It provides different levels of annotations for the check-out images. Comparing with the existing datasets, ours is closer to the realistic setting and can derive a variety of research problems. Besides the dataset, we also benchmark the performance on this dataset with various approaches. The dataset and related resources can be found at this https URL .",
"Contemporary Vision and Pattern Recognition problems such as face recognition, fingerprinting identification, image categorization, and DNA sequencing often have an arbitrarily large number of classes and properties to consider. To deal with such complex problems using just one feature descriptor is a difficult task and feature fusion may become mandatory. Although normal feature fusion is quite effective for some problems, it can yield unexpected classification results when the different features are not properly normalized and preprocessed. Besides it has the drawback of increasing the dimensionality which might require more training data. To cope with these problems, this paper introduces a unified approach that can combine many features and classifiers that requires less training and is more adequate to some problems than a naive method, where all features are simply concatenated and fed independently to each classification algorithm. Besides that, the presented technique is amenable to continuous learning, both when refining a learned model and also when adding new classes to be discriminated. The introduced fusion approach is validated using a multi-class fruit-and-vegetable categorization task in a semi-controlled environment, such as a distribution center or the supermarket cashier. The results show that the solution is able to reduce the classification error in up to 15 percentage points with respect to the baseline.",
"",
"We introduce the Densely Segmented Supermarket (D2S) dataset, a novel benchmark for instance-aware semantic segmentation in an industrial domain. It contains 21 000 high-resolution images with pixel-wise labels of all object instances. The objects comprise groceries and everyday products from 60 categories. The benchmark is designed such that it resembles the real-world setting of an automatic checkout, inventory, or warehouse system. The training images only contain objects of a single class on a homogeneous background, while the validation and test sets are much more complex and diverse. To further benchmark the robustness of instance segmentation methods, the scenes are acquired with different lightings, rotations, and backgrounds. We ensure that there are no ambiguities in the labels and that every instance is labeled comprehensively. The annotations are pixel-precise and allow using crops of single instances for articial data augmentation. The dataset covers several challenges highly relevant in the field, such as a limited amount of training data and a high diversity in the test and validation sets. The evaluation of state-of-the-art object detection and instance segmentation methods on D2S reveals significant room for improvement."
]
} |
1904.04866 | 2939874124 | Analysis of word embedding properties to inform their use in downstream NLP tasks has largely been studied by assessing nearest neighbors. However, geometric properties of the continuous feature space contribute directly to the use of embedding features in downstream models, and are largely unexplored. We consider four properties of word embedding geometry, namely: position relative to the origin, distribution of features in the vector space, global pairwise distances, and local pairwise distances. We define a sequence of transformations to generate new embeddings that expose subsets of these properties to downstream models and evaluate change in task performance to understand the contribution of each property to NLP models. We transform publicly available pretrained embeddings from three popular toolkits (word2vec, GloVe, and FastText) and evaluate on a variety of intrinsic tasks, which model linguistic information in the vector space, and extrinsic tasks, which use vectors as input to machine learning models. We find that intrinsic evaluations are highly sensitive to absolute position, while extrinsic tasks rely primarily on local similarity. Our findings suggest that future embedding models and post-processing techniques should focus primarily on similarity to nearby points in vector space. | Word embedding models and outputs have been analyzed from several angles. In terms of performance, evaluating the quality'' of word embedding models has long been a thorny problem. While intrinsic evaluations such as word similarity and analogy completion are intuitive and easy to compute, they are limited by both confounding geometric factors @cite_16 and task-specific factors @cite_28 @cite_0 . show that these tasks, while correlated with some semantic content, do not always predict downstream performance. Thus, it is necessary to use a more comprehensive set of intrinsic and extrinsic evaluations for embeddings. | {
"cite_N": [
"@cite_28",
"@cite_16",
"@cite_0"
],
"mid": [
"2963419157",
"2963176474",
""
],
"abstract": [
"Lacking standardized extrinsic evaluation methods for vector representations of words, the NLP community has relied heavily on word similarity tasks as a proxy for intrinsic evaluation of word vectors. Word similarity evaluation, which correlates the distance between vectors and human judgments of “semantic similarity” is attractive, because it is computationally inexpensive and fast. In this paper we present several problems associated with the evaluation of word vectors on word similarity datasets, and summarize existing solutions. Our study suggests that the use of word similarity tasks for evaluation of word vectors is not sustainable and calls for further research on evaluation methods.",
"The offset method for solving word analo- gies has become a standard evaluation tool for vector-space semantic models: it is considered desirable for a space to repre- sent semantic relations as consistent vec- tor offsets. We show that the method's re- liance on cosine similarity conflates offset consistency with largely irrelevant neigh- borhood structure, and propose simple baselines that should be used to improve the utility of the method in vector space evaluation.",
""
]
} |
1904.04998 | 2935920407 | We present a novel method for simultaneous learning of depth, egomotion, object motion, and camera intrinsics from monocular videos, using only consistency across neighboring video frames as supervision signal. Similarly to prior work, our method learns by applying differentiable warping to frames and comparing the result to adjacent ones, but it provides several improvements: We address occlusions geometrically and differentiably, directly using the depth maps as predicted during training. We introduce randomized layer normalization, a novel powerful regularizer, and we account for object motion relative to the scene. To the best of our knowledge, our work is the first to learn the camera intrinsic parameters, including lens distortion, from video in an unsupervised manner, thereby allowing us to extract accurate depth and motion from arbitrary videos of unknown origin at scale. We evaluate our results on the Cityscapes, KITTI and EuRoC datasets, establishing new state of the art on depth prediction and odometry, and demonstrate qualitatively that depth prediction can be learned from a collection of YouTube videos. | Estimating scene depth is an important task for robot navigation and manipulation and historically much research has been devoted to it including large body of research on stereo, multi-view geometry, active sensing and so on @cite_14 @cite_19 @cite_16 . More recently learning-based approaches for depth prediction have taken center stage based on learning of dense prediction @cite_16 @cite_35 @cite_28 @cite_6 . In these, scene depth is predicted from input RGB images and the depth estimation function is learned using supervision provided by a sensor, such as a LiDAR. Similar approach is used for other dense predictions e.g. surface normals @cite_29 @cite_51 . | {
"cite_N": [
"@cite_35",
"@cite_14",
"@cite_28",
"@cite_29",
"@cite_6",
"@cite_19",
"@cite_16",
"@cite_51"
],
"mid": [
"1803059841",
"127013308",
"2963591054",
"1905829557",
"",
"2125416623",
"2171740948",
"1899309388"
],
"abstract": [
"In this article, we tackle the problem of depth estimation from single monocular images. Compared with depth estimation using multiple images such as stereo depth perception, depth from monocular images is much more challenging. Prior work typically focuses on exploiting geometric priors or additional sources of information, most using hand-crafted features. Recently, there is mounting evidence that features from deep convolutional neural networks (CNN) set new records for various vision applications. On the other hand, considering the continuous characteristic of the depth values, depth estimation can be naturally formulated as a continuous conditional random field (CRF) learning problem. Therefore, here we present a deep convolutional neural field model for estimating depths from single monocular images, aiming to jointly explore the capacity of deep CNN and continuous CRF. In particular, we propose a deep structured learning scheme which learns the unary and pairwise potentials of continuous CRF in a unified deep CNN framework. We then further propose an equally effective model based on fully convolutional networks and a novel superpixel pooling method, which is about 10 times faster, to speedup the patch-wise convolutions in the deep model. With this more efficient model, we are able to design deeper networks to pursue better performance. Our proposed method can be used for depth estimation of general scenes with no geometric priors nor any extra information injected. In our case, the integral of the partition function can be calculated in a closed form such that we can exactly solve the log-likelihood maximization. Moreover, solving the inference problem for predicting depths of a test image is highly efficient as closed-form solutions exist. Experiments on both indoor and outdoor scene datasets demonstrate that the proposed method outperforms state-of-the-art depth estimation approaches.",
"Depth estimation in computer vision and robotics is most commonly done via stereo vision (stereopsis), in which images from two cameras are used to triangulate and estimate distances. However, there are also numerous monocular visual cues--such as texture variations and gradients, defocus, color haze, etc. --that have heretofore been little exploited in such systems. Some of these cues apply even in regions without texture, where stereo would work poorly. In this paper, we apply a Markov Random Field (MRF) learning algorithm to capture some of these monocular cues, and incorporate them into a stereo system. We show that by adding monocular cues to stereo (triangulation) ones, we obtain significantly more accurate depth estimates than is possible using either monocular or stereo cues alone. This holds true for a large variety of environments, including both indoor environments and unstructured outdoor environments containing trees forests, buildings, etc. Our approach is general, and applies to incorporating monocular cues together with any off-the-shelf stereo system.",
"This paper addresses the problem of estimating the depth map of a scene given a single RGB image. We propose a fully convolutional architecture, encompassing residual learning, to model the ambiguous mapping between monocular images and depth maps. In order to improve the output resolution, we present a novel way to efficiently learn feature map up-sampling within the network. For optimization, we introduce the reverse Huber loss that is particularly suited for the task at hand and driven by the value distributions commonly present in depth maps. Our model is composed of a single architecture that is trained end-to-end and does not rely on post-processing techniques, such as CRFs or other additional refinement steps. As a result, it runs in real-time on images or videos. In the evaluation, we show that the proposed model contains fewer parameters and requires fewer training data than the current state of the art, while outperforming all approaches on depth estimation. Code and models are publicly available.",
"In this paper we address three different computer vision tasks using a single basic architecture: depth prediction, surface normal estimation, and semantic labeling. We use a multiscale convolutional network that is able to adapt easily to each task using only small modifications, regressing from the input image to the output map directly. Our method progressively refines predictions using a sequence of scales, and captures many image details without any superpixels or low-level segmentation. We achieve state-of-the-art performance on benchmarks for all three tasks.",
"",
"We consider the problem of depth estimation from a single monocular image in this work. It is a challenging task as no reliable depth cues are available, e.g., stereo correspondences, motions etc. Previous efforts have been focusing on exploiting geometric priors or additional sources of information, with all using hand-crafted features. Recently, there is mounting evidence that features from deep convolutional neural networks (CNN) are setting new records for various vision applications. On the other hand, considering the continuous characteristic of the depth values, depth estimations can be naturally formulated into a continuous conditional random field (CRF) learning problem. Therefore, we in this paper present a deep convolutional neural field model for estimating depths from a single image, aiming to jointly explore the capacity of deep CNN and continuous CRF. Specifically, we propose a deep structured learning scheme which learns the unary and pairwise potentials of continuous CRF in a unified deep CNN framework. The proposed method can be used for depth estimations of general scenes with no geometric priors nor any extra information injected. In our case, the integral of the partition function can be analytically calculated, thus we can exactly solve the log-likelihood optimization. Moreover, solving the MAP problem for predicting depths of a new image is highly efficient as closed-form solutions exist. We experimentally demonstrate that the proposed method outperforms state-of-the-art depth estimation methods on both indoor and outdoor scene datasets.",
"Predicting depth is an essential component in understanding the 3D geometry of a scene. While for stereo images local correspondence suffices for estimation, finding depth relations from a single image is less straightforward, requiring integration of both global and local information from various cues. Moreover, the task is inherently ambiguous, with a large source of uncertainty coming from the overall scale. In this paper, we present a new method that addresses this task by employing two deep network stacks: one that makes a coarse global prediction based on the entire image, and another that refines this prediction locally. We also apply a scale-invariant error to help measure depth relations rather than scale. By leveraging the raw datasets as large sources of training data, our method achieves state-of-the-art results on both NYU Depth and KITTI, and matches detailed depth boundaries without the need for superpixelation.",
"In the past few years, convolutional neural nets (CNN) have shown incredible promise for learning visual representations. In this paper, we use CNNs for the task of predicting surface normals from a single image. But what is the right architecture? We propose to build upon the decades of hard work in 3D scene understanding to design a new CNN architecture for the task of surface normal estimation. We show that incorporating several constraints (man-made, Manhattan world) and meaningful intermediate representations (room layout, edge labels) in the architecture leads to state of the art performance on surface normal estimation. We also show that our network is quite robust and show state of the art results on other datasets as well without any fine-tuning."
]
} |
1904.04998 | 2935920407 | We present a novel method for simultaneous learning of depth, egomotion, object motion, and camera intrinsics from monocular videos, using only consistency across neighboring video frames as supervision signal. Similarly to prior work, our method learns by applying differentiable warping to frames and comparing the result to adjacent ones, but it provides several improvements: We address occlusions geometrically and differentiably, directly using the depth maps as predicted during training. We introduce randomized layer normalization, a novel powerful regularizer, and we account for object motion relative to the scene. To the best of our knowledge, our work is the first to learn the camera intrinsic parameters, including lens distortion, from video in an unsupervised manner, thereby allowing us to extract accurate depth and motion from arbitrary videos of unknown origin at scale. We evaluate our results on the Cityscapes, KITTI and EuRoC datasets, establishing new state of the art on depth prediction and odometry, and demonstrate qualitatively that depth prediction can be learned from a collection of YouTube videos. | Unsupervised learning of depth, where the only supervision is obtained from the monocular video itself and no depth sensors are needed, has also been popularized recently @cite_46 @cite_2 @cite_45 @cite_3 @cite_9 @cite_36 @cite_7 . @cite_2 introduced joint learning of depth and ego-motion. @cite_46 demonstrated a fully differentiable such approach where depth and ego-motion are predicted jointly by deep neural networks. Techniques were developed for the monocular @cite_3 @cite_27 @cite_9 @cite_52 @cite_7 @cite_39 @cite_42 and binocular @cite_45 @cite_3 @cite_33 @cite_23 @cite_39 @cite_23 settings. It was shown that depth quality at inference is improved when stereo inputs are only used during training, unlike methods that relied on stereo disparity at inference too @cite_30 @cite_10 @cite_37 . Other novel techniques include the use of motion @cite_52 @cite_39 @cite_7 @cite_42 @cite_33 | {
"cite_N": [
"@cite_30",
"@cite_37",
"@cite_33",
"@cite_7",
"@cite_36",
"@cite_9",
"@cite_42",
"@cite_52",
"@cite_3",
"@cite_39",
"@cite_27",
"@cite_45",
"@cite_23",
"@cite_2",
"@cite_46",
"@cite_10"
],
"mid": [
"2604231069",
"2962793285",
"2810554001",
"",
"2608018946",
"2963906250",
"",
"2963549785",
"2561074213",
"2963654727",
"2767981409",
"2520707372",
"2962816904",
"2949634581",
"2609883120",
"2883040756"
],
"abstract": [
"We propose a novel deep learning architecture for regressing disparity from a rectified pair of stereo images. We leverage knowledge of the problem’s geometry to form a cost volume using deep feature representations. We learn to incorporate contextual information using 3-D convolutions over this volume. Disparity values are regressed from the cost volume using a proposed differentiable soft argmin operation, which allows us to train our method end-to-end to sub-pixel accuracy without any additional post-processing or regularization. We evaluate our method on the Scene Flow and KITTI datasets and on KITTI we set a new stateof-the-art benchmark, while being significantly faster than competing approaches.",
"We present an end-to-end deep learning architecture for depth map inference from multi-view images. In the network, we first extract deep visual image features, and then build the 3D cost volume upon the reference camera frustum via the differentiable homography warping. Next, we apply 3D convolutions to regularize and regress the initial depth map, which is then refined with the reference image to generate the final output. Our framework flexibly adapts arbitrary N-view inputs using a variance-based cost metric that maps multiple features into one cost feature. The proposed MVSNet is demonstrated on the large-scale indoor DTU dataset. With simple post-processing, our method not only significantly outperforms previous state-of-the-arts, but also is several times faster in runtime. We also evaluate MVSNet on the complex outdoor Tanks and Temples dataset, where our method ranks first before April 18, 2018 without any fine-tuning, showing the strong generalization ability of MVSNet.",
"Learning to estimate 3D geometry in a single image by watching unlabeled videos via deep convolutional network has made significant process recently. Current state-of-the-art (SOTA) methods, are based on the learning framework of rigid structure-from-motion, where only 3D camera ego motion is modeled for geometry estimation. However, moving objects also exist in many videos, e.g. moving cars in a street scene. In this paper, we tackle such motion by additionally incorporating per-pixel 3D object motion into the learning framework, which provides holistic 3D scene flow understanding and helps single image geometry estimation. Specifically, given two consecutive frames from a video, we adopt a motion network to predict their relative 3D camera pose and a segmentation mask distinguishing moving objects and rigid background. An optical flow network is used to estimate dense 2D per-pixel correspondence. A single image depth network predicts depth maps for both images. The four types of information, i.e. 2D flow, camera pose, segment mask and depth maps, are integrated into a differentiable holistic 3D motion parser (HMP), where per-pixel 3D motion for rigid background and moving objects are recovered. We design various losses w.r.t. the two types of 3D motions for training the depth and motion networks, yielding further error reduction for estimated geometry. Finally, in order to solve the 3D motion confusion from monocular videos, we combine stereo images into joint training. Experiments on KITTI 2015 dataset show that our estimated geometry, 3D motion and moving object masks, not only are constrained to be consistent, but also significantly outperforms other SOTA algorithms, demonstrating the benefits of our approach.",
"",
"We propose SfM-Net, a geometry-aware neural network for motion estimation in videos that decomposes frame-to-frame pixel motion in terms of scene and object depth, camera motion and 3D object rotations and translations. Given a sequence of frames, SfM-Net predicts depth, segmentation, camera and rigid object motions, converts those into a dense frame-to-frame motion field (optical flow), differentiably warps frames in time to match pixels and back-propagates. The model can be trained with various degrees of supervision: 1) self-supervised by the re-projection photometric error (completely unsupervised), 2) supervised by ego-motion (camera motion), or 3) supervised by depth (e.g., as provided by RGBD sensors). SfM-Net extracts meaningful depth estimates and successfully estimates frame-to-frame camera rotations and translations. It often successfully segments the moving objects in the scene, even though such supervision is never provided.",
"We present a novel approach for unsupervised learning of depth and ego-motion from monocular video. Unsupervised learning removes the need for separate supervisory signals (depth or ego-motion ground truth, or multi-view video). Prior work in unsupervised depth learning uses pixel-wise or gradient-based losses, which only consider pixels in small local neighborhoods. Our main contribution is to explicitly consider the inferred 3D geometry of the whole scene, and enforce consistency of the estimated 3D point clouds and ego-motion across consecutive frames. This is a challenging task and is solved by a novel (approximate) backpropagation algorithm for aligning 3D structures. We combine this novel 3D-based loss with 2D losses based on photometric quality of frame reconstructions using estimated depth and ego-motion from adjacent frames. We also incorporate validity masks to avoid penalizing areas in which no useful information exists. We test our algorithm on the KITTI dataset and on a video dataset captured on an uncalibrated mobile phone camera. Our proposed approach consistently improves depth estimates on both datasets, and outperforms the state-of-the-art for both depth and ego-motion. Because we only require a simple video, learning depth and ego-motion on large and varied datasets becomes possible. We demonstrate this by training on the low quality uncalibrated video dataset and evaluating on KITTI, ranking among top performing prior methods which are trained on KITTI itself.1",
"",
"Learning to estimate 3D geometry in a single image by watching unlabeled videos via deep convolutional network is attracting significant attention. In this paper, we introduce a \"3D as-smooth-as-possible (3D-ASAP)\" prior inside the pipeline, which enables joint estimation of edges and 3D scene, yielding results with significant improvement in accuracy for fine detailed structures. Specifically, we define the 3D-ASAP prior by requiring that any two points recovered in 3D from an image should lie on an existing planar surface if no other cues provided. We design an unsupervised framework that Learns Edges and Geometry (depth, normal) all at Once (LEGO). The predicted edges are embedded into depth and surface normal smoothness terms, where pixels without edges in-between are constrained to satisfy the prior. In our framework, the predicted depths, normals and edges are forced to be consistent all the time. We conduct experiments on KITTI to evaluate our estimated geometry and CityScapes to perform edge evaluation. We show that in all of the tasks, i.e. depth, normal and edge, our algorithm vastly outperforms other state-of-the-art (SOTA) algorithms, demonstrating the benefits of our approach.",
"In this paper we formulate structure from motion as a learning problem. We train a convolutional network end-to-end to compute depth and camera motion from successive, unconstrained image pairs. The architecture is composed of multiple stacked encoder-decoder networks, the core part being an iterative network that is able to improve its own predictions. The network estimates not only depth and motion, but additionally surface normals, optical flow between the images and confidence of the matching. A crucial component of the approach is a training loss based on spatial relative differences. Compared to traditional two-frame structure from motion methods, results are more accurate and more robust. In contrast to the popular depth-from-single-image networks, DeMoN learns the concept of matching and, thus, better generalizes to structures not seen during training.",
"The ability to predict depth from a single image - using recent advances in CNNs - is of increasing interest to the vision community. Unsupervised strategies to learning are particularly appealing as they can utilize much larger and varied monocular video datasets during learning without the need for ground truth depth or stereo. In previous works, separate pose and depth CNN predictors had to be determined such that their joint outputs minimized the photometric error. Inspired by recent advances in direct visual odometry (DVO), we argue that the depth CNN predictor can be learned without a pose CNN predictor. Further, we demonstrate empirically that incorporation of a differentiable implementation of DVO, along with a novel depth normalization strategy - substantially improves performance over state of the art that use monocular videos for training.",
"Learning to reconstruct depths in a single image by watching unlabeled videos via deep convolutional network (DCN) is attracting significant attention in recent years. In this paper, we introduce a surface normal representation for unsupervised depth estimation framework. Our estimated depths are constrained to be compatible with predicted normals, yielding more robust geometry results. Specifically, we formulate an edge-aware depth-normal consistency term, and solve it by constructing a depth-to-normal layer and a normal-to-depth layer inside of the DCN. The depth-to-normal layer takes estimated depths as input, and computes normal directions using cross production based on neighboring pixels. Then given the estimated normals, the normal-to-depth layer outputs a regularized depth map through local planar smoothness. Both layers are computed with awareness of edges inside the image to help address the issue of depth normal discontinuity and preserve sharp edges. Finally, to train the network, we apply the photometric error and gradient smoothness for both depth and normal predictions. We conducted experiments on both outdoor (KITTI) and indoor (NYUv2) datasets, and show that our algorithm vastly outperforms state of the art, which demonstrates the benefits from our approach.",
"Learning based methods have shown very promising results for the task of depth estimation in single images. However, most existing approaches treat depth prediction as a supervised regression problem and as a result, require vast quantities of corresponding ground truth depth data for training. Just recording quality depth data in a range of environments is a challenging problem. In this paper, we innovate beyond existing approaches, replacing the use of explicit depth data during training with easier-to-obtain binocular stereo footage. We propose a novel training objective that enables our convolutional neural network to learn to perform single image depth estimation, despite the absence of ground truth depth data. Ex-ploiting epipolar geometry constraints, we generate disparity images by training our network with an image reconstruction loss. We show that solving for image reconstruction alone results in poor quality depth images. To overcome this problem, we propose a novel training loss that enforces consistency between the disparities produced relative to both the left and right images, leading to improved performance and robustness compared to existing approaches. Our method produces state of the art results for monocular depth estimation on the KITTI driving dataset, even outperforming supervised methods that have been trained with ground truth depth.",
"Despite learning based methods showing promising results in single view depth estimation and visual odometry, most existing approaches treat the tasks in a supervised manner. Recent approaches to single view depth estimation explore the possibility of learning without full supervision via minimizing photometric error. In this paper, we explore the use of stereo sequences for learning depth and visual odometry. The use of stereo sequences enables the use of both spatial (between left-right pairs) and temporal (forward backward) photometric warp error, and constrains the scene depth and camera motion to be in a common, real-world scale. At test time our framework is able to estimate single view depth and two-view odometry from a monocular sequence. We also show how we can improve on a standard photometric warp loss by considering a warp of deep features. We show through extensive experiments that: (i) jointly training for single view depth and visual odometry improves depth prediction because of the additional constraint imposed on depths and achieves competitive results for visual odometry; (ii) deep feature-based warping loss improves upon simple photometric warp loss for both single view depth estimation and visual odometry. Our method outperforms existing learning based methods on the KITTI driving dataset in both tasks. The source code is available at https: github.com Huangying-Zhan Depth-VO-Feat.",
"A significant weakness of most current deep Convolutional Neural Networks is the need to train them using vast amounts of manu- ally labelled data. In this work we propose a unsupervised framework to learn a deep convolutional neural network for single view depth predic- tion, without requiring a pre-training stage or annotated ground truth depths. We achieve this by training the network in a manner analogous to an autoencoder. At training time we consider a pair of images, source and target, with small, known camera motion between the two such as a stereo pair. We train the convolutional encoder for the task of predicting the depth map for the source image. To do so, we explicitly generate an inverse warp of the target image using the predicted depth and known inter-view displacement, to reconstruct the source image; the photomet- ric error in the reconstruction is the reconstruction loss for the encoder. The acquisition of this training data is considerably simpler than for equivalent systems, requiring no manual annotation, nor calibration of depth sensor to camera. We show that our network trained on less than half of the KITTI dataset (without any further augmentation) gives com- parable performance to that of the state of art supervised methods for single view depth estimation.",
"We present an unsupervised learning framework for the task of monocular depth and camera motion estimation from unstructured video sequences. In common with recent work [10, 14, 16], we use an end-to-end learning approach with view synthesis as the supervisory signal. In contrast to the previous work, our method is completely unsupervised, requiring only monocular video sequences for training. Our method uses single-view depth and multiview pose networks, with a loss based on warping nearby views to the target using the computed depth and pose. The networks are thus coupled by the loss during training, but can be applied independently at test time. Empirical evaluation on the KITTI dataset demonstrates the effectiveness of our approach: 1) monocular depth performs comparably with supervised methods that use either ground-truth pose or depth for training, and 2) pose estimation performs favorably compared to established SLAM systems under comparable input settings.",
"This paper presents StereoNet, the first end-to-end deep architecture for real-time stereo matching that runs at 60fps on an NVidia Titan X, producing high-quality, edge-preserved, quantization-free disparity maps. A key insight of this paper is that the network achieves a sub-pixel matching precision than is a magnitude higher than those of traditional stereo matching approaches. This allows us to achieve real-time performance by using a very low resolution cost volume that encodes all the information needed to achieve high disparity precision. Spatial precision is achieved by employing a learned edge-aware upsampling function. Our model uses a Siamese network to extract features from the left and right image. A first estimate of the disparity is computed in a very low resolution cost volume, then hierarchically the model re-introduces high-frequency details through a learned upsampling function that uses compact pixel-to-pixel refinement networks. Leveraging color input as a guide, this function is capable of producing high-quality edge-aware output. We achieve compelling results on multiple benchmarks, showing how the proposed method offers extreme flexibility at an acceptable computational budget."
]
} |
1904.04998 | 2935920407 | We present a novel method for simultaneous learning of depth, egomotion, object motion, and camera intrinsics from monocular videos, using only consistency across neighboring video frames as supervision signal. Similarly to prior work, our method learns by applying differentiable warping to frames and comparing the result to adjacent ones, but it provides several improvements: We address occlusions geometrically and differentiably, directly using the depth maps as predicted during training. We introduce randomized layer normalization, a novel powerful regularizer, and we account for object motion relative to the scene. To the best of our knowledge, our work is the first to learn the camera intrinsic parameters, including lens distortion, from video in an unsupervised manner, thereby allowing us to extract accurate depth and motion from arbitrary videos of unknown origin at scale. We evaluate our results on the Cityscapes, KITTI and EuRoC datasets, establishing new state of the art on depth prediction and odometry, and demonstrate qualitatively that depth prediction can be learned from a collection of YouTube videos. | Learning depth from images in the wild is also an active research field, mostly focusing on single or multi-view images @cite_25 @cite_44 @cite_48 . It is especially hard as shown by for internet photos, due to the diversity of input sources and no knowledge of the camera parameters @cite_48 . Our work makes a step in addressing this challenge by learning intrinsics for videos in the wild. | {
"cite_N": [
"@cite_44",
"@cite_48",
"@cite_25"
],
"mid": [
"2471962767",
"2963760790",
"2536680313"
],
"abstract": [
"Incremental Structure-from-Motion is a prevalent strategy for 3D reconstruction from unordered image collections. While incremental reconstruction systems have tremendously advanced in all regards, robustness, accuracy, completeness, and scalability remain the key problems towards building a truly general-purpose pipeline. We propose a new SfM technique that improves upon the state of the art to make a further step towards this ultimate goal. The full reconstruction pipeline is released to the public as an open-source implementation.",
"Single-view depth prediction is a fundamental problem in computer vision. Recently, deep learning methods have led to significant progress, but such methods are limited by the available training data. Current datasets based on 3D sensors have key limitations, including indoor-only images (NYU), small numbers of training examples (Make3D), and sparse sampling (KITTI). We propose to use multi-view Internet photo collections, a virtually unlimited data source, to generate training data via modern structure-from-motion and multi-view stereo (MVS) methods, and present a large depth dataset called MegaDepth based on this idea. Data derived from MVS comes with its own challenges, including noise and unreconstructable objects. We address these challenges with new data cleaning methods, as well as automatically augmenting our data with ordinal depth relations generated using semantic segmentation. We validate the use of large amounts of Internet data by showing that models trained on MegaDepth exhibit strong generalization-not only to novel scenes, but also to other diverse datasets including Make3D, KITTI, and DIW, even when no images from those datasets are seen during training.1",
"We present a system that can match and reconstruct 3D scenes from extremely large collections of photographs such as those found by searching for a given city (e.g., Rome) on Internet photo sharing sites. Our system uses a collection of novel parallel distributed matching and reconstruction algorithms, designed to maximize parallelism at each stage in the pipeline and minimize serialization bottlenecks. It is designed to scale gracefully with both the size of the problem and the amount of available computation. We have experimented with a variety of alternative algorithms at each stage of the pipeline and report on which ones work best in a parallel computing environment. Our experimental results demonstrate that it is now possible to reconstruct cities consisting of 150K images in less than a day on a cluster with 500 compute cores."
]
} |
1904.04998 | 2935920407 | We present a novel method for simultaneous learning of depth, egomotion, object motion, and camera intrinsics from monocular videos, using only consistency across neighboring video frames as supervision signal. Similarly to prior work, our method learns by applying differentiable warping to frames and comparing the result to adjacent ones, but it provides several improvements: We address occlusions geometrically and differentiably, directly using the depth maps as predicted during training. We introduce randomized layer normalization, a novel powerful regularizer, and we account for object motion relative to the scene. To the best of our knowledge, our work is the first to learn the camera intrinsic parameters, including lens distortion, from video in an unsupervised manner, thereby allowing us to extract accurate depth and motion from arbitrary videos of unknown origin at scale. We evaluate our results on the Cityscapes, KITTI and EuRoC datasets, establishing new state of the art on depth prediction and odometry, and demonstrate qualitatively that depth prediction can be learned from a collection of YouTube videos. | Multiple approaches have been proposed for handling occlusions in the context of optical flow @cite_31 @cite_38 @cite_22 . These approaches are disconnected from geometry. Differentiable mesh rendering, has recently been attracting increasing attention @cite_41 @cite_26 , begins to adopt a geometric approach to occlusions. In the context of learning to predict depth and egmotion, occlusions were addressed via a learned explainability mask @cite_46 , by penalizing the minimum reprojection loss between the previous frame or the next frame into the middle one, and through optical flow @cite_33 . In the context of unsupervised learning of depth from video, we are the first to propose a direct geometric approach to occlusions via a differentiable loss. | {
"cite_N": [
"@cite_38",
"@cite_26",
"@cite_22",
"@cite_33",
"@cite_41",
"@cite_46",
"@cite_31"
],
"mid": [
"2894983388",
"",
"2962905430",
"2810554001",
"2963327506",
"2609883120",
"2962864875"
],
"abstract": [
"Learning optical flow with neural networks is hampered by the need for obtaining training data with associated ground truth. Unsupervised learning is a promising direction, yet the performance of current unsupervised methods is still limited. In particular, the lack of proper occlusion handling in commonly used data terms constitutes a major source of error. While most optical flow methods process pairs of consecutive frames, more advanced occlusion reasoning can be realized when considering multiple frames. In this paper, we propose a framework for unsupervised learning of optical flow and occlusions over multiple frames. More specifically, we exploit the minimal configuration of three frames to strengthen the photometric loss and explicitly reason about occlusions. We demonstrate that our multi-frame, occlusion-sensitive formulation outperforms existing unsupervised two-frame methods and even produces results on par with some fully supervised methods.",
"",
"Two optical flow estimation problems are addressed: (i) occlusion estimation and handling, and (ii) estimation from image sequences longer than two frames. The proposed ContinualFlow method estimates occlusions before flow, avoiding the use of flow corrupted by occlusions for their estimation. We show that providing occlusion masks as an additional input to flow estimation improves the standard performance metric by more than 25 on both KITTI and Sintel. As a second contribution, a novel method for incorporating information from past frames into flow estimation is introduced. The previous frame flow serves as an input to occlusion estimation and as a prior in occluded regions, i.e. those without visual correspondences. By continually using the previous frame flow, ContinualFlow performance improves further by 18 on KITTI and 7 on Sintel, achieving top performance on KITTI and Sintel.",
"Learning to estimate 3D geometry in a single image by watching unlabeled videos via deep convolutional network has made significant process recently. Current state-of-the-art (SOTA) methods, are based on the learning framework of rigid structure-from-motion, where only 3D camera ego motion is modeled for geometry estimation. However, moving objects also exist in many videos, e.g. moving cars in a street scene. In this paper, we tackle such motion by additionally incorporating per-pixel 3D object motion into the learning framework, which provides holistic 3D scene flow understanding and helps single image geometry estimation. Specifically, given two consecutive frames from a video, we adopt a motion network to predict their relative 3D camera pose and a segmentation mask distinguishing moving objects and rigid background. An optical flow network is used to estimate dense 2D per-pixel correspondence. A single image depth network predicts depth maps for both images. The four types of information, i.e. 2D flow, camera pose, segment mask and depth maps, are integrated into a differentiable holistic 3D motion parser (HMP), where per-pixel 3D motion for rigid background and moving objects are recovered. We design various losses w.r.t. the two types of 3D motions for training the depth and motion networks, yielding further error reduction for estimated geometry. Finally, in order to solve the 3D motion confusion from monocular videos, we combine stereo images into joint training. Experiments on KITTI 2015 dataset show that our estimated geometry, 3D motion and moving object masks, not only are constrained to be consistent, but also significantly outperforms other SOTA algorithms, demonstrating the benefits of our approach.",
"Traditional computer graphics rendering pipeline is designed for procedurally generating 2D quality images from 3D shapes with high performance. The non-differentiability due to discrete operations such as visibility computation makes it hard to explicitly correlate rendering parameters and the resulting image, posing a significant challenge for inverse rendering tasks. Recent work on differentiable rendering achieves differentiability either by designing surrogate gradients for non-differentiable operations or via an approximate but differentiable renderer. These methods, however, are still limited when it comes to handling occlusion and restricted to particular rendering effects. We present RenderNet, a differentiable rendering convolutional network with a novel projection unit that can render 2D images from 3D shapes. Spatial occlusion and shading calculation are automatically encoded in the network. Our experiments show that RenderNet can successfully learn to implement different shaders, and can be used in inverse rendering tasks to estimate shape, pose, lighting and texture from a single image.",
"We present an unsupervised learning framework for the task of monocular depth and camera motion estimation from unstructured video sequences. In common with recent work [10, 14, 16], we use an end-to-end learning approach with view synthesis as the supervisory signal. In contrast to the previous work, our method is completely unsupervised, requiring only monocular video sequences for training. Our method uses single-view depth and multiview pose networks, with a loss based on warping nearby views to the target using the computed depth and pose. The networks are thus coupled by the loss during training, but can be applied independently at test time. Empirical evaluation on the KITTI dataset demonstrates the effectiveness of our approach: 1) monocular depth performs comparably with supervised methods that use either ground-truth pose or depth for training, and 2) pose estimation performs favorably compared to established SLAM systems under comparable input settings.",
"It has been recently shown that a convolutional neural network can learn optical flow estimation with unsupervised learning. However, the performance of the unsupervised methods still has a relatively large gap compared to its supervised counterpart. Occlusion and large motion are some of the major factors that limit the current unsupervised learning of optical flow methods. In this work we introduce a new method which models occlusion explicitly and a new warping way that facilitates the learning of large motion. Our method shows promising results on Flying Chairs, MPI-Sintel and KITTI benchmark datasets. Especially on KITTI dataset where abundant unlabeled samples exist, our unsupervised method outperforms its counterpart trained with supervised learning."
]
} |
1904.04998 | 2935920407 | We present a novel method for simultaneous learning of depth, egomotion, object motion, and camera intrinsics from monocular videos, using only consistency across neighboring video frames as supervision signal. Similarly to prior work, our method learns by applying differentiable warping to frames and comparing the result to adjacent ones, but it provides several improvements: We address occlusions geometrically and differentiably, directly using the depth maps as predicted during training. We introduce randomized layer normalization, a novel powerful regularizer, and we account for object motion relative to the scene. To the best of our knowledge, our work is the first to learn the camera intrinsic parameters, including lens distortion, from video in an unsupervised manner, thereby allowing us to extract accurate depth and motion from arbitrary videos of unknown origin at scale. We evaluate our results on the Cityscapes, KITTI and EuRoC datasets, establishing new state of the art on depth prediction and odometry, and demonstrate qualitatively that depth prediction can be learned from a collection of YouTube videos. | Learning to predict the camera intrinsics has mostly been limited to strongly supervised approaches. The sources of groundtruth varies: @cite_12 use focal lengths estimated employing classical 1D structure from motion. @cite_13 obtain the focal length based on EXIF. @cite_0 synthesize images from panoramas using virtual cameras with known intrinsics, including distortion. To our knowledge, our approach is the only one that learns the camera intrinsics in an unsupervised manner, directly from video, jointly with depth, ego-motion and object motion. | {
"cite_N": [
"@cite_0",
"@cite_13",
"@cite_12"
],
"mid": [
"2903345161",
"2624218322",
"2294109271"
],
"abstract": [
"Calibration of wide field-of-view cameras is a fundamental step for numerous visual media production applications, such as 3D reconstruction, image undistortion, augmented reality and camera motion estimation. However, existing calibration methods require multiple images of a calibration pattern (typically a checkerboard), assume the presence of lines, require manual interaction and or need an image sequence. In contrast, we present a novel fully automatic deep learning-based approach that overcomes all these limitations and works with a single image of general scenes. Our approach builds upon the recent developments in deep Convolutional Neural Networks (CNN): our network automatically estimates the intrinsic parameters of the camera (focal length and distortion parameter) from a single input image. In order to train the CNN, we leverage the great amount of omnidirectional images available on the Internet to automatically generate a large-scale dataset composed of millions of wide field-of-view images with ground truth intrinsic parameters. Experiments successfully demonstrated the quality of our results, both quantitatively and qualitatively.",
"The focal length information of an image is indispensable for many computer vision tasks. In general, focal length can be obtained via camera calibration using specific planner patterns. However, for images taken by an unknown device, focal length can only be estimated based on the image itself. Currently, most of the single-image focal length estimation methods make use of predefined geometric cues (such as vanishing points or parallel lines) to infer focal length, which constrains their applications mainly on manmade scenes. The machine learning algorithms have demonstrated great performance in many computer vision tasks, but these methods are seldom used in the focal length estimation task, partially due to the shortage of labeled images for training the model. To bridge this gap, we first introduce a large-scale dataset FocaLens, which is especially designed for single-image focal length estimation. Taking advantage of the FocaLens dataset, we also propose a new focal length estimation model, which exploits the multiscale detection architecture to encode object distributions in images to assist focal length estimation. Additionally, an online focal transformation approach is proposed to further promote the model’s generalization ability. Experimental results demonstrate that the proposed model trained on FocaLens can not only achieve state-of-the-art results on the scenes with distinct geometric cues but also obtain comparable results on the scenes even without distinct geometric cues.",
""
]
} |
1904.05020 | 2939566018 | Person re-identification (Re-ID) models usually show a limited performance when they are trained on one dataset and tested on another dataset due to the inter-dataset bias (e.g. completely different identities and backgrounds) and the intra-dataset difference (e.g. camera invariance). In terms of this issue, given a labelled source training set and an unlabelled target training set, we propose an unsupervised transfer learning method characterized by 1) bridging inter-dataset bias and intra-dataset difference via a proposed ImitateModel simultaneously; 2) regarding the unsupervised person Re-ID problem as a semi-supervised learning problem formulated by a dual classification loss to learn a discriminative representation across domains; 3) exploiting the underlying commonality across different domains from the class-style space to improve the generalization ability of re-ID models. Extensive experiments are conducted on two widely employed benchmarks, including Market-1501 and DukeMTMC-reID, and experimental results demonstrate that the proposed method can achieve a competitive performance against other state-of-the-art unsupervised Re-ID approaches. | Supervised person re-ID. Most existing person re-ID models are based on supervised learning, trained on a large number of labelled images across cameras. They focus on feature engineering @cite_32 @cite_20 @cite_22 @cite_29 @cite_54 , distance metric learning @cite_44 @cite_48 @cite_4 @cite_5 @cite_37 , or creating new deep learning architectures @cite_21 @cite_30 . For example, Kalayeh al @cite_33 learned both part-level features and global features. Chen al @cite_44 proposed a quadruplet loss to handle the weakness of the triplet loss on person re-ID. Li al @cite_6 proposed a new person re-ID network with a joint learning of soft and hard attentions, which took advantage of both joint learning attention selection and feature representation. Although these models offer a promising performance on recent person Re-ID datasets ( , Market-1501 @cite_1 and DukeMTMC-reID @cite_36 @cite_46 ), it is hard to utilize in practical applications due to the demand of tremendous labelled data. | {
"cite_N": [
"@cite_30",
"@cite_37",
"@cite_4",
"@cite_22",
"@cite_33",
"@cite_36",
"@cite_48",
"@cite_54",
"@cite_29",
"@cite_21",
"@cite_32",
"@cite_6",
"@cite_1",
"@cite_44",
"@cite_5",
"@cite_46",
"@cite_20"
],
"mid": [
"",
"",
"",
"",
"",
"2951254243",
"",
"",
"",
"1928419358",
"1518138188",
"2962926870",
"2204750386",
"2606377603",
"",
"",
""
],
"abstract": [
"",
"",
"",
"",
"",
"To help accelerate progress in multi-target, multi-camera tracking systems, we present (i) a new pair of precision-recall measures of performance that treats errors of all types uniformly and emphasizes correct identification over sources of error; (ii) the largest fully-annotated and calibrated data set to date with more than 2 million frames of 1080p, 60fps video taken by 8 cameras observing more than 2,700 identities over 85 minutes; and (iii) a reference software system as a comparison baseline. We show that (i) our measures properly account for bottom-line identity match performance in the multi-camera setting; (ii) our data set poses realistic challenges to current trackers; and (iii) the performance of our system is comparable to the state of the art.",
"",
"",
"",
"In this work, we propose a method for simultaneously learning features and a corresponding similarity metric for person re-identification. We present a deep convolutional architecture with layers specially designed to address the problem of re-identification. Given a pair of images as input, our network outputs a similarity value indicating whether the two input images depict the same person. Novel elements of our architecture include a layer that computes cross-input neighborhood differences, which capture local relationships between the two input images based on mid-level features from each input image. A high-level summary of the outputs of this layer is computed by a layer of patch summary features, which are then spatially integrated in subsequent layers. Our method significantly outperforms the state of the art on both a large data set (CUHK03) and a medium-sized data set (CUHK01), and is resistant to over-fitting. We also demonstrate that by initially training on an unrelated large data set before fine-tuning on a small target data set, our network can achieve results comparable to the state of the art even on a small data set (VIPeR).",
"Viewpoint invariant pedestrian recognition is an important yet under-addressed problem in computer vision. This is likely due to the difficulty in matching two objects with unknown viewpoint and pose. This paper presents a method of performing viewpoint invariant pedestrian recognition using an efficiently and intelligently designed object representation, the ensemble of localized features (ELF). Instead of designing a specific feature by hand to solve the problem, we define a feature space using our intuition about the problem and let a machine learning algorithm find the best representation. We show how both an object class specific representation and a discriminative recognition model can be learned using the AdaBoost algorithm. This approach allows many different kinds of simple features to be combined into a single similarity function. The method is evaluated using a viewpoint invariant pedestrian recognition dataset and the results are shown to be superior to all previous benchmarks for both recognition and reacquisition of pedestrians.",
"Existing person re-identification (re-id) methods either assume the availability of well-aligned person bounding box images as model input or rely on constrained attention selection mechanisms to calibrate misaligned images. They are therefore sub-optimal for re-id matching in arbitrarily aligned person images potentially with large human pose variations and unconstrained auto-detection errors. In this work, we show the advantages of jointly learning attention selection and feature representation in a Convolutional Neural Network (CNN) by maximising the complementary information of different levels of visual attention subject to re-id discriminative learning constraints. Specifically, we formulate a novel Harmonious Attention CNN (HA-CNN) model for joint learning of soft pixel attention and hard regional attention along with simultaneous optimisation of feature representations, dedicated to optimise person re-id in uncontrolled (misaligned) images. Extensive comparative evaluations validate the superiority of this new HA-CNN model for person re-id over a wide variety of state-of-the-art methods on three large-scale benchmarks including CUHK03, Market-1501, and DukeMTMC-ReID.",
"This paper contributes a new high quality dataset for person re-identification, named \"Market-1501\". Generally, current datasets: 1) are limited in scale, 2) consist of hand-drawn bboxes, which are unavailable under realistic settings, 3) have only one ground truth and one query image for each identity (close environment). To tackle these problems, the proposed Market-1501 dataset is featured in three aspects. First, it contains over 32,000 annotated bboxes, plus a distractor set of over 500K images, making it the largest person re-id dataset to date. Second, images in Market-1501 dataset are produced using the Deformable Part Model (DPM) as pedestrian detector. Third, our dataset is collected in an open system, where each identity has multiple images under each camera. As a minor contribution, inspired by recent advances in large-scale image search, this paper proposes an unsupervised Bag-of-Words descriptor. We view person re-identification as a special task of image search. In experiment, we show that the proposed descriptor yields competitive accuracy on VIPeR, CUHK03, and Market-1501 datasets, and is scalable on the large-scale 500k dataset.",
"Person re-identification (ReID) is an important task in wide area video surveillance which focuses on identifying people across different cameras. Recently, deep learning networks with a triplet loss become a common framework for person ReID. However, the triplet loss pays main attentions on obtaining correct orders on the training set. It still suffers from a weaker generalization capability from the training set to the testing set, thus resulting in inferior performance. In this paper, we design a quadruplet loss, which can lead to the model output with a larger inter-class variation and a smaller intra-class variation compared to the triplet loss. As a result, our model has a better generalization ability and can achieve a higher performance on the testing set. In particular, a quadruplet deep network using a margin-based online hard negative mining is proposed based on the quadruplet loss for the person ReID. In extensive experiments, the proposed network outperforms most of the state-of-the-art algorithms on representative datasets which clearly demonstrates the effectiveness of our proposed method.",
"",
"",
""
]
} |
1904.04514 | 2940262938 | High-resolution representation learning plays an essential role in many vision problems, e.g., pose estimation and semantic segmentation. The high-resolution network (HRNet) SunXLW19 , recently developed for human pose estimation, maintains high-resolution representations through the whole process by connecting high-to-low resolution convolutions in and produces strong high-resolution representations by repeatedly conducting fusions across parallel convolutions. In this paper, we conduct a further study on high-resolution representations by introducing a simple yet effective modification and apply it to a wide range of vision tasks. We augment the high-resolution representation by aggregating the (upsampled) representations from all the parallel convolutions rather than only the representation from the high-resolution convolution as done in SunXLW19 . This simple modification leads to stronger representations, evidenced by superior results. We show top results in semantic segmentation on Cityscapes, LIP, and PASCAL Context, and facial landmark detection on AFLW, COFW, @math W, and WFLW. In addition, we build a multi-level representation from the high-resolution representation and apply it to the Faster R-CNN object detection framework and the extended frameworks. The proposed approach achieves superior results to existing single-model networks on COCO object detection. The code and models have been publicly available at this https URL . | Strong high-resolution representations play an essential role in pixel and region labeling problems, e.g., semantic segmentation, human pose estimation, facial landmark detection, and object detection. We review representation learning techniques developed mainly in the semantic segmentation, facial landmark detection @cite_67 @cite_68 @cite_74 @cite_92 @cite_136 @cite_95 @cite_64 and object detection areas The techniques developed for human pose estimation are reviewed in @cite_5 . , from low-resolution representation learning, high-resolution representation recovering, to high-resolution representation maintaining. | {
"cite_N": [
"@cite_67",
"@cite_64",
"@cite_136",
"@cite_92",
"@cite_95",
"@cite_74",
"@cite_5",
"@cite_68"
],
"mid": [
"1976948919",
"2166694921",
"1896424170",
"2519753233",
"2474575620",
"2740020909",
"2916798096",
"2623892189"
],
"abstract": [
"We propose a new approach for estimation of the positions of facial key points with three-level carefully designed convolutional networks. At each level, the outputs of multiple networks are fused for robust and accurate estimation. Thanks to the deep structures of convolutional networks, global high-level features are extracted over the whole face region at the initialization stage, which help to locate high accuracy key points. There are two folds of advantage for this. First, the texture context information over the entire face is utilized to locate each key point. Second, since the networks are trained to predict all the key points simultaneously, the geometric constraints among key points are implicitly encoded. The method therefore can avoid local minimum caused by ambiguity and data corruption in difficult image samples due to occlusions, large pose variations, and extreme lightings. The networks at the following two levels are trained to locally refine initial predictions and their inputs are limited to small regions around the initial predictions. Several network structures critical for accurate and robust facial point detection are investigated. Extensive experiments show that our approach outperforms state-of-the-art methods in both detection accuracy and reliability.",
"Accurate face alignment is a vital prerequisite step for most face perception tasks such as face recognition, facial expression analysis and non-realistic face re-rendering. It can be formulated as the nonlinear inference of the facial landmarks from the detected face region. Deep network seems a good choice to model the nonlinearity, but it is nontrivial to apply it directly. In this paper, instead of a straightforward application of deep network, we propose a Coarse-to-Fine Auto-encoder Networks (CFAN) approach, which cascades a few successive Stacked Auto-encoder Networks (SANs). Specifically, the first SAN predicts the landmarks quickly but accurately enough as a preliminary, by taking as input a low-resolution version of the detected face holistically. The following SANs then progressively refine the landmark by taking as input the local features extracted around the current landmarks (output of the previous SAN) with higher and higher resolution. Extensive experiments conducted on three challenging datasets demonstrate that our CFAN outperforms the state-of-the-art methods and performs in real-time(40+fps excluding face detection on a desktop).",
"Facial landmark detection has long been impeded by the problems of occlusion and pose variation. Instead of treating the detection task as a single and independent problem, we investigate the possibility of improving detection robustness through multi-task learning. Specifically, we wish to optimize facial landmark detection together with heterogeneous but subtly correlated tasks, e.g. head pose estimation and facial attribute inference. This is non-trivial since different tasks have different learning difficulties and convergence rates. To address this problem, we formulate a novel tasks-constrained deep model, with task-wise early stopping to facilitate learning convergence. Extensive evaluations show that the proposed task-constrained learning (i) outperforms existing methods, especially in dealing with faces with severe occlusion and pose variation, and (ii) reduces model complexity drastically compared to the state-of-the-art method based on cascaded deep model [21].",
"In this work, we introduce a novel Recurrent Attentive-Refinement (RAR) network for facial landmark detection under unconstrained conditions, suffering from challenges like facial occlusions and or pose variations. RAR follows the pipeline of cascaded regressions that refines landmark locations progressively. However, instead of updating all the landmark locations together, RAR refines the landmark locations sequentially at each recurrent stage. In this way, more reliable landmark points are refined earlier and help to infer locations of other challenging landmarks that may stay with occlusions and or extreme poses. RAR can thus effectively control detection errors from those challenging landmarks and improve overall performance even in presence of heavy occlusions and or extreme conditions. To determine the sequence of landmarks, RAR employs an attentive-refinement mechanism. The attention LSTM (A-LSTM) and refinement LSTM (R-LSTM) models are introduced in RAR. At each recurrent stage, A-LSTM implicitly identifies a reliable landmark as the attention center. Following the sequence of attention centers, R-LSTM sequentially refines the landmarks near or correlated with the attention centers and provides ultimate detection results finally. To further enhance algorithmic robustness, instead of using mean shape for initialization, RAR adaptively determines the initialization by selecting from a pool of shape centers clustered from all training shapes. As an end-to-end trainable model, RAR demonstrates superior performance in detecting challenging landmarks in comprehensive experiments and it also establishes new state-of-the-arts on the 300-W, COFW and AFLW benchmark datasets.",
"Cascaded regression has recently become the method of choice for solving non-linear least squares problems such as deformable image alignment. Given a sizeable training set, cascaded regression learns a set of generic rules that are sequentially applied to minimise the least squares problem. Despite the success of cascaded regression for problems such as face alignment and head pose estimation, there are several shortcomings arising in the strategies proposed thus far. Specifically, (a) the regressors are learnt independently, (b) the descent directions may cancel one another out and (c) handcrafted features (e.g., HoGs, SIFT etc.) are mainly used to drive the cascade, which may be sub-optimal for the task at hand. In this paper, we propose a combined and jointly trained convolutional recurrent neural network architecture that allows the training of an end-to-end to system that attempts to alleviate the aforementioned drawbacks. The recurrent module facilitates the joint optimisation of the regressors by assuming the cascades form a nonlinear dynamical system, in effect fully utilising the information between all cascade levels by introducing a memory unit that shares information across all levels. The convolutional module allows the network to extract features that are specialised for the task at hand and are experimentally shown to outperform hand-crafted features. We show that the application of the proposed architecture for the problem of face alignment results in a strong improvement over the current state-of-the-art.",
"Regression based facial landmark detection methods usually learns a series of regression functions to update the landmark positions from an initial estimation. Most of existing approaches focus on learning effective mapping functions with robust image features to improve performance. The approach to dealing with the initialization issue, however, receives relatively fewer attentions. In this paper, we present a deep regression architecture with two-stage re-initialization to explicitly deal with the initialization problem. At the global stage, given an image with a rough face detection result, the full face region is firstly re-initialized by a supervised spatial transformer network to a canonical shape state and then trained to regress a coarse landmark estimation. At the local stage, different face parts are further separately re-initialized to their own canonical shape states, followed by another regression subnetwork to get the final estimation. Our proposed deep architecture is trained from end to end and obtains promising results using different kinds of unstable initialization. It also achieves superior performances over many competing algorithms.",
"",
"In this paper, we propose Deep Alignment Network (DAN), a robust face alignment method based on a deep neural network architecture. DAN consists of multiple stages, where each stage improves the locations of the facial landmarks estimated by the previous stage. Our method uses entire face images at all stages, contrary to the recently proposed face alignment methods that rely on local patches. This is possible thanks to the use of landmark heatmaps which provide visual information about landmark locations estimated at the previous stages of the algorithm. The use of entire face images rather than patches allows DAN to handle face images with large variation in head pose and difficult initializations. An extensive evaluation on two publicly available datasets shows that DAN reduces the state-of-the-art failure rate by up to 70 . Our method has also been submitted for evaluation as part of the Menpo challenge."
]
} |
1904.04514 | 2940262938 | High-resolution representation learning plays an essential role in many vision problems, e.g., pose estimation and semantic segmentation. The high-resolution network (HRNet) SunXLW19 , recently developed for human pose estimation, maintains high-resolution representations through the whole process by connecting high-to-low resolution convolutions in and produces strong high-resolution representations by repeatedly conducting fusions across parallel convolutions. In this paper, we conduct a further study on high-resolution representations by introducing a simple yet effective modification and apply it to a wide range of vision tasks. We augment the high-resolution representation by aggregating the (upsampled) representations from all the parallel convolutions rather than only the representation from the high-resolution convolution as done in SunXLW19 . This simple modification leads to stronger representations, evidenced by superior results. We show top results in semantic segmentation on Cityscapes, LIP, and PASCAL Context, and facial landmark detection on AFLW, COFW, @math W, and WFLW. In addition, we build a multi-level representation from the high-resolution representation and apply it to the Faster R-CNN object detection framework and the extended frameworks. The proposed approach achieves superior results to existing single-model networks on COCO object detection. The code and models have been publicly available at this https URL . | The fully-convolutional network (FCN) approaches @cite_127 @cite_90 compute low-resolution representations by removing the fully-connected layers in a classification network, and estimate from their coarse segmentation confidence maps. The estimated segmentation maps are improved by combining the fine segmentation score maps estimated from intermediate low-level medium-resolution representations @cite_127 , or iterating the processes @cite_68 . Similar techniques have also been applied to edge detection, e.g., holistic edge detection @cite_65 . | {
"cite_N": [
"@cite_90",
"@cite_127",
"@cite_68",
"@cite_65"
],
"mid": [
"1487583988",
"",
"2623892189",
"2950891598"
],
"abstract": [
"We present an integrated framework for using Convolutional Networks for classification, localization and detection. We show how a multiscale and sliding window approach can be efficiently implemented within a ConvNet. We also introduce a novel deep learning approach to localization by learning to predict object boundaries. Bounding boxes are then accumulated rather than suppressed in order to increase detection confidence. We show that different tasks can be learned simultaneously using a single shared network. This integrated framework is the winner of the localization task of the ImageNet Large Scale Visual Recognition Challenge 2013 (ILSVRC2013) and obtained very competitive results for the detection and classifications tasks. In post-competition work, we establish a new state of the art for the detection task. Finally, we release a feature extractor from our best model called OverFeat.",
"",
"In this paper, we propose Deep Alignment Network (DAN), a robust face alignment method based on a deep neural network architecture. DAN consists of multiple stages, where each stage improves the locations of the facial landmarks estimated by the previous stage. Our method uses entire face images at all stages, contrary to the recently proposed face alignment methods that rely on local patches. This is possible thanks to the use of landmark heatmaps which provide visual information about landmark locations estimated at the previous stages of the algorithm. The use of entire face images rather than patches allows DAN to handle face images with large variation in head pose and difficult initializations. An extensive evaluation on two publicly available datasets shows that DAN reduces the state-of-the-art failure rate by up to 70 . Our method has also been submitted for evaluation as part of the Menpo challenge.",
"We develop a new edge detection algorithm that tackles two important issues in this long-standing vision problem: (1) holistic image training and prediction; and (2) multi-scale and multi-level feature learning. Our proposed method, holistically-nested edge detection (HED), performs image-to-image prediction by means of a deep learning model that leverages fully convolutional neural networks and deeply-supervised nets. HED automatically learns rich hierarchical representations (guided by deep supervision on side responses) that are important in order to approach the human ability resolve the challenging ambiguity in edge and object boundary detection. We significantly advance the state-of-the-art on the BSD500 dataset (ODS F-score of .782) and the NYU Depth dataset (ODS F-score of .746), and do so with an improved speed (0.4 second per image) that is orders of magnitude faster than some recent CNN-based edge detection algorithms."
]
} |
1904.04514 | 2940262938 | High-resolution representation learning plays an essential role in many vision problems, e.g., pose estimation and semantic segmentation. The high-resolution network (HRNet) SunXLW19 , recently developed for human pose estimation, maintains high-resolution representations through the whole process by connecting high-to-low resolution convolutions in and produces strong high-resolution representations by repeatedly conducting fusions across parallel convolutions. In this paper, we conduct a further study on high-resolution representations by introducing a simple yet effective modification and apply it to a wide range of vision tasks. We augment the high-resolution representation by aggregating the (upsampled) representations from all the parallel convolutions rather than only the representation from the high-resolution convolution as done in SunXLW19 . This simple modification leads to stronger representations, evidenced by superior results. We show top results in semantic segmentation on Cityscapes, LIP, and PASCAL Context, and facial landmark detection on AFLW, COFW, @math W, and WFLW. In addition, we build a multi-level representation from the high-resolution representation and apply it to the Faster R-CNN object detection framework and the extended frameworks. The proposed approach achieves superior results to existing single-model networks on COCO object detection. The code and models have been publicly available at this https URL . | The fully convolutional network is extended, by replacing a few (typically two) strided convolutions and the associated convolutions with dilated convolutions, to the dilation version, leading to medium-resolution representations @cite_72 @cite_86 @cite_9 @cite_44 @cite_60 . The representations are further augmented to multi-scale contextual representations @cite_72 @cite_86 @cite_118 through feature pyramids for segmenting objects at multiple scales. | {
"cite_N": [
"@cite_118",
"@cite_60",
"@cite_9",
"@cite_44",
"@cite_72",
"@cite_86"
],
"mid": [
"2962891704",
"",
"",
"2964288706",
"2560023338",
"2412782625"
],
"abstract": [
"Incorporating multi-scale features in fully convolutional neural networks (FCNs) has been a key element to achieving state-of-the-art performance on semantic image segmentation. One common way to extract multi-scale features is to feed multiple resized input images to a shared deep network and then merge the resulting features for pixelwise classification. In this work, we propose an attention mechanism that learns to softly weight the multi-scale features at each pixel location. We adapt a state-of-the-art semantic image segmentation model, which we jointly train with multi-scale input images and the attention model. The proposed attention model not only outperforms averageand max-pooling, but allows us to diagnostically visualize the importance of features at different positions and scales. Moreover, we show that adding extra supervision to the output at each scale is essential to achieving excellent performance when merging multi-scale features. We demonstrate the effectiveness of our model with extensive experiments on three challenging datasets, including PASCAL-Person-Part, PASCAL VOC 2012 and a subset of MS-COCO 2014.",
"",
"",
"Abstract: Deep Convolutional Neural Networks (DCNNs) have recently shown state of the art performance in high level vision tasks, such as image classification and object detection. This work brings together methods from DCNNs and probabilistic graphical models for addressing the task of pixel-level classification (also called \"semantic image segmentation\"). We show that responses at the final layer of DCNNs are not sufficiently localized for accurate object segmentation. This is due to the very invariance properties that make DCNNs good for high level tasks. We overcome this poor localization property of deep networks by combining the responses at the final DCNN layer with a fully connected Conditional Random Field (CRF). Qualitatively, our \"DeepLab\" system is able to localize segment boundaries at a level of accuracy which is beyond previous methods. Quantitatively, our method sets the new state-of-art at the PASCAL VOC-2012 semantic image segmentation task, reaching 71.6 IOU accuracy in the test set. We show how these results can be obtained efficiently: Careful network re-purposing and a novel application of the 'hole' algorithm from the wavelet community allow dense computation of neural net responses at 8 frames per second on a modern GPU.",
"Scene parsing is challenging for unrestricted open vocabulary and diverse scenes. In this paper, we exploit the capability of global context information by different-region-based context aggregation through our pyramid pooling module together with the proposed pyramid scene parsing network (PSPNet). Our global prior representation is effective to produce good quality results on the scene parsing task, while PSPNet provides a superior framework for pixel-level prediction. The proposed approach achieves state-of-the-art performance on various datasets. It came first in ImageNet scene parsing challenge 2016, PASCAL VOC 2012 benchmark and Cityscapes benchmark. A single PSPNet yields the new record of mIoU accuracy 85.4 on PASCAL VOC 2012 and accuracy 80.2 on Cityscapes.",
"In this work we address the task of semantic image segmentation with Deep Learning and make three main contributions that are experimentally shown to have substantial practical merit. First , we highlight convolution with upsampled filters, or ‘atrous convolution’, as a powerful tool in dense prediction tasks. Atrous convolution allows us to explicitly control the resolution at which feature responses are computed within Deep Convolutional Neural Networks. It also allows us to effectively enlarge the field of view of filters to incorporate larger context without increasing the number of parameters or the amount of computation. Second , we propose atrous spatial pyramid pooling (ASPP) to robustly segment objects at multiple scales. ASPP probes an incoming convolutional feature layer with filters at multiple sampling rates and effective fields-of-views, thus capturing objects as well as image context at multiple scales. Third , we improve the localization of object boundaries by combining methods from DCNNs and probabilistic graphical models. The commonly deployed combination of max-pooling and downsampling in DCNNs achieves invariance but has a toll on localization accuracy. We overcome this by combining the responses at the final DCNN layer with a fully connected Conditional Random Field (CRF), which is shown both qualitatively and quantitatively to improve localization performance. Our proposed “DeepLab” system sets the new state-of-art at the PASCAL VOC-2012 semantic image segmentation task, reaching 79.7 percent mIOU in the test set, and advances the results on three other datasets: PASCAL-Context, PASCAL-Person-Part, and Cityscapes. All of our code is made publicly available online."
]
} |
1904.04514 | 2940262938 | High-resolution representation learning plays an essential role in many vision problems, e.g., pose estimation and semantic segmentation. The high-resolution network (HRNet) SunXLW19 , recently developed for human pose estimation, maintains high-resolution representations through the whole process by connecting high-to-low resolution convolutions in and produces strong high-resolution representations by repeatedly conducting fusions across parallel convolutions. In this paper, we conduct a further study on high-resolution representations by introducing a simple yet effective modification and apply it to a wide range of vision tasks. We augment the high-resolution representation by aggregating the (upsampled) representations from all the parallel convolutions rather than only the representation from the high-resolution convolution as done in SunXLW19 . This simple modification leads to stronger representations, evidenced by superior results. We show top results in semantic segmentation on Cityscapes, LIP, and PASCAL Context, and facial landmark detection on AFLW, COFW, @math W, and WFLW. In addition, we build a multi-level representation from the high-resolution representation and apply it to the Faster R-CNN object detection framework and the extended frameworks. The proposed approach achieves superior results to existing single-model networks on COCO object detection. The code and models have been publicly available at this https URL . | An upsample subnetwork, like a decoder, is adopted to gradually recover the high-resolution representations from the low-resolution representations outputted by the downsample process. The upsample subnetwork could be a symmetric version of the downsample subnetwork, with skipping connection over some mirrored layers to transform the pooling indices, e.g., SegNet @cite_81 and DeconvNet @cite_11 , or copying the feature maps, e.g., U-Net @cite_114 and Hourglass @cite_124 @cite_51 @cite_130 @cite_88 @cite_80 , encoder-decoder @cite_78 , FPN @cite_38 , and so on. The full-resolution residual network @cite_42 introduces an extra full-resolution stream that carries information at the full image resolution, to replace the skip connections, and each unit in the downsample and upsample subnetworks receives information from and sends information to the full-resolution stream. | {
"cite_N": [
"@cite_38",
"@cite_78",
"@cite_42",
"@cite_130",
"@cite_81",
"@cite_88",
"@cite_80",
"@cite_124",
"@cite_51",
"@cite_114",
"@cite_11"
],
"mid": [
"2565639579",
"2963544488",
"2950668883",
"",
"2963881378",
"",
"",
"2307770531",
"",
"1901129140",
"1745334888"
],
"abstract": [
"Feature pyramids are a basic component in recognition systems for detecting objects at different scales. But pyramid representations have been avoided in recent object detectors that are based on deep convolutional networks, partially because they are slow to compute and memory intensive. In this paper, we exploit the inherent multi-scale, pyramidal hierarchy of deep convolutional networks to construct feature pyramids with marginal extra cost. A top-down architecture with lateral connections is developed for building high-level semantic feature maps at all scales. This architecture, called a Feature Pyramid Network (FPN), shows significant improvement as a generic feature extractor in several applications. Using a basic Faster R-CNN system, our method achieves state-of-the-art single-model results on the COCO detection benchmark without bells and whistles, surpassing all existing single-model entries including those from the COCO 2016 challenge winners. In addition, our method can run at 5 FPS on a GPU and thus is a practical and accurate solution to multi-scale object detection. Code will be made publicly available.",
"We propose a novel recurrent encoder-decoder network model for real-time video-based face alignment. Our proposed model predicts 2D facial point maps regularized by a regression loss, while uniquely exploiting recurrent learning at both spatial and temporal dimensions. At the spatial level, we add a feedback loop connection between the combined output response map and the input, in order to enable iterative coarse-to-fine face alignment using a single network model. At the temporal level, we first decouple the features in the bottleneck of the network into temporal-variant factors, such as pose and expression, and temporal-invariant factors, such as identity information. Temporal recurrent learning is then applied to the decoupled temporal-variant features, yielding better generalization and significantly more accurate results at test time. We perform a comprehensive experimental analysis, showing the importance of each component of our proposed model, as well as superior results over the state-of-the-art in standard datasets.",
"Semantic image segmentation is an essential component of modern autonomous driving systems, as an accurate understanding of the surrounding scene is crucial to navigation and action planning. Current state-of-the-art approaches in semantic image segmentation rely on pre-trained networks that were initially developed for classifying images as a whole. While these networks exhibit outstanding recognition performance (i.e., what is visible?), they lack localization accuracy (i.e., where precisely is something located?). Therefore, additional processing steps have to be performed in order to obtain pixel-accurate segmentation masks at the full image resolution. To alleviate this problem we propose a novel ResNet-like architecture that exhibits strong localization and recognition performance. We combine multi-scale context with pixel-level accuracy by using two processing streams within our network: One stream carries information at the full image resolution, enabling precise adherence to segment boundaries. The other stream undergoes a sequence of pooling operations to obtain robust features for recognition. The two streams are coupled at the full image resolution using residuals. Without additional processing steps and without pre-training, our approach achieves an intersection-over-union score of 71.8 on the Cityscapes dataset.",
"",
"We present a novel and practical deep fully convolutional neural network architecture for semantic pixel-wise segmentation termed SegNet. This core trainable segmentation engine consists of an encoder network, a corresponding decoder network followed by a pixel-wise classification layer. The architecture of the encoder network is topologically identical to the 13 convolutional layers in the VGG16 network [1] . The role of the decoder network is to map the low resolution encoder feature maps to full input resolution feature maps for pixel-wise classification. The novelty of SegNet lies is in the manner in which the decoder upsamples its lower resolution input feature map(s). Specifically, the decoder uses pooling indices computed in the max-pooling step of the corresponding encoder to perform non-linear upsampling. This eliminates the need for learning to upsample. The upsampled maps are sparse and are then convolved with trainable filters to produce dense feature maps. We compare our proposed architecture with the widely adopted FCN [2] and also with the well known DeepLab-LargeFOV [3] , DeconvNet [4] architectures. This comparison reveals the memory versus accuracy trade-off involved in achieving good segmentation performance. SegNet was primarily motivated by scene understanding applications. Hence, it is designed to be efficient both in terms of memory and computational time during inference. It is also significantly smaller in the number of trainable parameters than other competing architectures and can be trained end-to-end using stochastic gradient descent. We also performed a controlled benchmark of SegNet and other architectures on both road scenes and SUN RGB-D indoor scene segmentation tasks. These quantitative assessments show that SegNet provides good performance with competitive inference time and most efficient inference memory-wise as compared to other architectures. We also provide a Caffe implementation of SegNet and a web demo at http: mi.eng.cam.ac.uk projects segnet .",
"",
"",
"This work introduces a novel convolutional network architecture for the task of human pose estimation. Features are processed across all scales and consolidated to best capture the various spatial relationships associated with the body. We show how repeated bottom-up, top-down processing used in conjunction with intermediate supervision is critical to improving the performance of the network. We refer to the architecture as a “stacked hourglass” network based on the successive steps of pooling and upsampling that are done to produce a final set of predictions. State-of-the-art results are achieved on the FLIC and MPII benchmarks outcompeting all recent methods.",
"",
"There is large consent that successful training of deep networks requires many thousand annotated training samples. In this paper, we present a network and training strategy that relies on the strong use of data augmentation to use the available annotated samples more efficiently. The architecture consists of a contracting path to capture context and a symmetric expanding path that enables precise localization. We show that such a network can be trained end-to-end from very few images and outperforms the prior best method (a sliding-window convolutional network) on the ISBI challenge for segmentation of neuronal structures in electron microscopic stacks. Using the same network trained on transmitted light microscopy images (phase contrast and DIC) we won the ISBI cell tracking challenge 2015 in these categories by a large margin. Moreover, the network is fast. Segmentation of a 512x512 image takes less than a second on a recent GPU. The full implementation (based on Caffe) and the trained networks are available at http: lmb.informatik.uni-freiburg.de people ronneber u-net .",
"We propose a novel semantic segmentation algorithm by learning a deep deconvolution network. We learn the network on top of the convolutional layers adopted from VGG 16-layer net. The deconvolution network is composed of deconvolution and unpooling layers, which identify pixelwise class labels and predict segmentation masks. We apply the trained network to each proposal in an input image, and construct the final semantic segmentation map by combining the results from all proposals in a simple manner. The proposed algorithm mitigates the limitations of the existing methods based on fully convolutional networks by integrating deep deconvolution network and proposal-wise prediction, our segmentation method typically identifies detailed structures and handles objects in multiple scales naturally. Our network demonstrates outstanding performance in PASCAL VOC 2012 dataset, and we achieve the best accuracy (72.5 ) among the methods trained without using Microsoft COCO dataset through ensemble with the fully convolutional network."
]
} |
1904.04514 | 2940262938 | High-resolution representation learning plays an essential role in many vision problems, e.g., pose estimation and semantic segmentation. The high-resolution network (HRNet) SunXLW19 , recently developed for human pose estimation, maintains high-resolution representations through the whole process by connecting high-to-low resolution convolutions in and produces strong high-resolution representations by repeatedly conducting fusions across parallel convolutions. In this paper, we conduct a further study on high-resolution representations by introducing a simple yet effective modification and apply it to a wide range of vision tasks. We augment the high-resolution representation by aggregating the (upsampled) representations from all the parallel convolutions rather than only the representation from the high-resolution convolution as done in SunXLW19 . This simple modification leads to stronger representations, evidenced by superior results. We show top results in semantic segmentation on Cityscapes, LIP, and PASCAL Context, and facial landmark detection on AFLW, COFW, @math W, and WFLW. In addition, we build a multi-level representation from the high-resolution representation and apply it to the Faster R-CNN object detection framework and the extended frameworks. The proposed approach achieves superior results to existing single-model networks on COCO object detection. The code and models have been publicly available at this https URL . | The asymmetric upsample process is also widely studied. RefineNet @cite_83 improves the combination of upsampled representations and the representations of the same resolution copied from the downsample process. Other works include: light upsample process @cite_20 ; light downsample and heavy upsample processes @cite_7 , recombinator networks @cite_32 ; improving skip connections with more or complicated convolutional units @cite_73 @cite_128 @cite_53 , as well as sending information from low-resolution skip connections to high-resolution skip connections @cite_70 or exchanging information between them @cite_103 ; studying the details the upsample process @cite_58 ; combining multi-scale pyramid representations @cite_104 @cite_84 ; stacking multiple DeconvNets U-Nets Hourglass @cite_126 @cite_121 with dense connections @cite_113 . | {
"cite_N": [
"@cite_128",
"@cite_7",
"@cite_70",
"@cite_103",
"@cite_53",
"@cite_58",
"@cite_32",
"@cite_84",
"@cite_104",
"@cite_121",
"@cite_113",
"@cite_83",
"@cite_126",
"@cite_73",
"@cite_20"
],
"mid": [
"",
"2897558462",
"2884436604",
"2892703603",
"",
"2738804062",
"2963544187",
"",
"2964309882",
"",
"2887748972",
"",
"2745943519",
"2598666589",
"2518965973"
],
"abstract": [
"",
"In this paper we present DCFE, a real-time facial landmark regression method based on a coarse-to-fine Ensemble of Regression Trees (ERT). We use a simple Convolutional Neural Network (CNN) to generate probability maps of landmarks location. These are further refined with the ERT regressor, which is initialized by fitting a 3D face model to the landmark maps. The coarse-to-fine structure of the ERT lets us address the combinatorial explosion of parts deformation. With the 3D model we also tackle other key problems such as robust regressor initialization, self occlusions, and simultaneous frontal and profile face analysis. In the experiments DCFE achieves the best reported result in AFLW, COFW, and 300 W private and common public data sets.",
"In this paper, we present UNet++, a new, more powerful architecture for medical image segmentation. Our architecture is essentially a deeply-supervised encoder-decoder network where the encoder and decoder sub-networks are connected through a series of nested, dense skip pathways. The re-designed skip pathways aim at reducing the semantic gap between the feature maps of the encoder and decoder sub-networks. We argue that the optimizer would deal with an easier learning task when the feature maps from the decoder and encoder networks are semantically similar. We have evaluated UNet++ in comparison with U-Net and wide U-Net architectures across multiple medical image segmentation tasks: nodule segmentation in the low-dose CT scans of chest, nuclei segmentation in the microscopy images, liver segmentation in abdominal CT scans, and polyp segmentation in colonoscopy videos. Our experiments demonstrate that UNet++ with deep supervision achieves an average IoU gain of 3.9 and 3.4 points over U-Net and wide U-Net, respectively.",
"",
"",
"Many machine vision applications require predictions for every pixel of the input image (for example semantic segmentation, boundary detection). Models for such problems usually consist of encoders which decreases spatial resolution while learning a high-dimensional representation, followed by decoders who recover the original input resolution and result in low-dimensional predictions. While encoders have been studied rigorously, relatively few studies address the decoder side. Therefore this paper presents an extensive comparison of a variety of decoders for a variety of pixel-wise prediction tasks. Our contributions are: (1) Decoders matter: we observe significant variance in results between different types of decoders on various problems. (2) We introduce a novel decoder: bilinear additive upsampling. (3) We introduce new residual-like connections for decoders. (4) We identify two decoder types which give a consistently high performance.",
"Deep neural networks with alternating convolutional, max-pooling and decimation layers are widely used in state of the art architectures for computer vision. Max-pooling purposefully discards precise spatial information in order to create features that are more robust, and typically organized as lower resolution spatial feature maps. On some tasks, such as whole-image classification, max-pooling derived features are well suited, however, for tasks requiring precise localization, such as pixel level prediction and segmentation, max-pooling destroys exactly the information required to perform well. Precise localization may be preserved by shallow convnets without pooling but at the expense of robustness. Can we have our max-pooled multilayered cake and eat it too? Several papers have proposed summation and concatenation based methods for combining upsampled coarse, abstract features with finer features to produce robust pixel level predictions. Here we introduce another model — dubbed Recombinator Networks — where coarse features inform finer features early in their formation such that finer features can make use of several layers of computation in deciding how to use coarse features. The model is trained once, end-to-end and performs better than summation-based architectures, reducing the error from the previous state of the art on two facial keypoint datasets, AFW and AFLW, by 30 and beating the current state-of-the-art on 300W without using extra data. We improve performance even further by adding a denoising prediction model based on a novel convnet formulation.",
"",
"Spatial pyramid pooling module or encode-decoder structure are used in deep neural networks for semantic segmentation task. The former networks are able to encode multi-scale contextual information by probing the incoming features with filters or pooling operations at multiple rates and multiple effective fields-of-view, while the latter networks can capture sharper object boundaries by gradually recovering the spatial information. In this work, we propose to combine the advantages from both methods. Specifically, our proposed model, DeepLabv3+, extends DeepLabv3 by adding a simple yet effective decoder module to refine the segmentation results especially along object boundaries. We further explore the Xception model and apply the depthwise separable convolution to both Atrous Spatial Pyramid Pooling and decoder modules, resulting in a faster and stronger encoder-decoder network. We demonstrate the effectiveness of the proposed model on PASCAL VOC 2012 and Cityscapes datasets, achieving the test set performance of 89 and 82.1 without any post-processing. Our paper is accompanied with a publicly available reference implementation of the proposed models in Tensorflow at https: github.com tensorflow models tree master research deeplab.",
"",
"In this paper, we propose quantized densely connected U-Nets for efficient visual landmark localization. The idea is that features of the same semantic meanings are globally reused across the stacked U-Nets. This dense connectivity largely improves the information flow, yielding improved localization accuracy. However, a vanilla dense design would suffer from critical efficiency issue in both training and testing. To solve this problem, we first propose order-K dense connectivity to trim off long-distance shortcuts; then, we use a memory-efficient implementation to significantly boost the training efficiency and investigate an iterative refinement that may slice the model size in half. Finally, to reduce the memory consumption and high precision operations both in training and testing, we further quantize weights, inputs, and gradients of our localization network to low bit-width numbers. We validate our approach in two tasks: human pose estimation and face alignment. The results show that our approach achieves state-of-the-art localization accuracy, but using ( )70 fewer parameters, ( )98 less model size and saving ( )32 ( ) training memory compared with other benchmark localizers.",
"",
"Recent progress in semantic segmentation has been driven by improving the spatial resolution under Fully Convolutional Networks (FCNs). To address this problem, we propose a Stacked Deconvolutional Network (SDN) for semantic segmentation. In SDN, multiple shallow deconvolutional networks, which are called as SDN units, are stacked one by one to integrate contextual information and guarantee the fine recovery of localization information. Meanwhile, inter-unit and intra-unit connections are designed to assist network training and enhance feature fusion since the connections improve the flow of information and gradient propagation throughout the network. Besides, hierarchical supervision is applied during the upsampling process of each SDN unit, which guarantees the discrimination of feature representations and benefits the network optimization. We carry out comprehensive experiments and achieve the new state-of-the-art results on three datasets, including PASCAL VOC 2012, CamVid, GATECH. In particular, our best model without CRF post-processing achieves an intersection-over-union score of 86.6 in the test set.",
"One of recent trends [31, 32, 14] in network architecture design is stacking small filters (e.g., 1x1 or 3x3) in the entire network because the stacked small filters is more efficient than a large kernel, given the same computational complexity. However, in the field of semantic segmentation, where we need to perform dense per-pixel prediction, we find that the large kernel (and effective receptive field) plays an important role when we have to perform the classification and localization tasks simultaneously. Following our design principle, we propose a Global Convolutional Network to address both the classification and localization issues for the semantic segmentation. We also suggest a residual-based boundary refinement to further refine the object boundaries. Our approach achieves state-of-art performance on two public benchmarks and significantly outperforms previous results, 82.2 (vs 80.2 ) on PASCAL VOC 2012 dataset and 76.9 (vs 71.8 ) on Cityscapes dataset.",
"This paper is on human pose estimation using Convolutional Neural Networks. Our main contribution is a CNN cascaded architecture specifically designed for learning part relationships and spatial context, and robustly inferring pose even for the case of severe part occlusions. To this end, we propose a detection-followed-by-regression CNN cascade. The first part of our cascade outputs part detection heatmaps and the second part performs regression on these heatmaps. The benefits of the proposed architecture are multi-fold: It guides the network where to focus in the image and effectively encodes part constraints and context. More importantly, it can effectively cope with occlusions because part detection heatmaps for occluded parts provide low confidence scores which subsequently guide the regression part of our network to rely on contextual information in order to predict the location of these parts. Additionally, we show that the proposed cascade is flexible enough to readily allow the integration of various CNN architectures for both detection and regression, including recent ones based on residual learning. Finally, we illustrate that our cascade achieves top performance on the MPII and LSP data sets. Code can be downloaded from http: www.cs.nott.ac.uk psxab5 ."
]
} |
1904.04514 | 2940262938 | High-resolution representation learning plays an essential role in many vision problems, e.g., pose estimation and semantic segmentation. The high-resolution network (HRNet) SunXLW19 , recently developed for human pose estimation, maintains high-resolution representations through the whole process by connecting high-to-low resolution convolutions in and produces strong high-resolution representations by repeatedly conducting fusions across parallel convolutions. In this paper, we conduct a further study on high-resolution representations by introducing a simple yet effective modification and apply it to a wide range of vision tasks. We augment the high-resolution representation by aggregating the (upsampled) representations from all the parallel convolutions rather than only the representation from the high-resolution convolution as done in SunXLW19 . This simple modification leads to stronger representations, evidenced by superior results. We show top results in semantic segmentation on Cityscapes, LIP, and PASCAL Context, and facial landmark detection on AFLW, COFW, @math W, and WFLW. In addition, we build a multi-level representation from the high-resolution representation and apply it to the Faster R-CNN object detection framework and the extended frameworks. The proposed approach achieves superior results to existing single-model networks on COCO object detection. The code and models have been publicly available at this https URL . | High-resolution representations are maintained through the whole process, typically by a network that is formed by connecting multi-resolution (from high-resolution to low-resolution) parallel convolutions with repeated information exchange across parallel convolutions. Representative works include GridNet @cite_89 , convolutional neural fabrics @cite_22 , interlinked CNNs @cite_55 , and the recently-developed high-resolution networks (HRNet) @cite_5 that is our interest. | {
"cite_N": [
"@cite_5",
"@cite_55",
"@cite_22",
"@cite_89"
],
"mid": [
"2916798096",
"2295744361",
"2964081403",
"2963741402"
],
"abstract": [
"",
"Face parsing is a basic task in face image analysis. It amounts to labeling each pixel with appropriate facial parts such as eyes and nose. In the paper, we present a interlinked convolutional neural network iCNN for solving this problem in an end-to-end fashion. It consists of multiple convolutional neural networks CNNs taking input in different scales. A special interlinking layer is designed to allow the CNNs to exchange information, enabling them to integrate local and contextual information efficiently. The hallmark of iCNN is the extensive use of downsampling and upsampling in the interlinking layers, while traditional CNNs usually uses downsampling only. A two-stage pipeline is proposed for face parsing and both stages use iCNN. The first stage localizes facial parts in the size-reduced image and the second stage labels the pixels in the identified facial parts in the original image. On a benchmark dataset we have obtained better results than the state-of-the-art methods.",
"Despite the success of CNNs, selecting the optimal architecture for a given task remains an open problem. Instead of aiming to select a single optimal architecture, we propose a \" fabric \" that embeds an exponentially large number of architectures. The fabric consists of a 3D trellis that connects response maps at different layers, scales, and channels with a sparse homogeneous local connectivity pattern. The only hyper-parameters of a fabric are the number of channels and layers. While individual architectures can be recovered as paths, the fabric can in addition ensemble all embedded architectures together, sharing their weights where their paths overlap. Parameters can be learned using standard methods based on back-propagation, at a cost that scales linearly in the fabric size. We present benchmark results competitive with the state of the art for image classification on MNIST and CIFAR10, and for semantic segmentation on the Part Labels dataset.",
"This paper presents GridNet, a new Convolutional Neural Network (CNN) architecture for semantic image segmentation (full scene labelling). Classical neural networks are implemented as one stream from the input to the output with subsampling operators applied in the stream in order to reduce the feature maps size and to increase the receptive field for the final prediction. However, for semantic image segmentation, where the task consists in providing a semantic class to each pixel of an image, feature maps reduction is harmful because it leads to a resolution loss in the output prediction. To tackle this problem, our GridNet follows a grid pattern allowing multiple interconnected streams to work at different resolutions. We show that our network generalizes many well known networks such as conv-deconv, residual or U-Net networks. GridNet is trained from scratch and achieves competitive results on the Cityscapes dataset."
]
} |
1904.04514 | 2940262938 | High-resolution representation learning plays an essential role in many vision problems, e.g., pose estimation and semantic segmentation. The high-resolution network (HRNet) SunXLW19 , recently developed for human pose estimation, maintains high-resolution representations through the whole process by connecting high-to-low resolution convolutions in and produces strong high-resolution representations by repeatedly conducting fusions across parallel convolutions. In this paper, we conduct a further study on high-resolution representations by introducing a simple yet effective modification and apply it to a wide range of vision tasks. We augment the high-resolution representation by aggregating the (upsampled) representations from all the parallel convolutions rather than only the representation from the high-resolution convolution as done in SunXLW19 . This simple modification leads to stronger representations, evidenced by superior results. We show top results in semantic segmentation on Cityscapes, LIP, and PASCAL Context, and facial landmark detection on AFLW, COFW, @math W, and WFLW. In addition, we build a multi-level representation from the high-resolution representation and apply it to the Faster R-CNN object detection framework and the extended frameworks. The proposed approach achieves superior results to existing single-model networks on COCO object detection. The code and models have been publicly available at this https URL . | The two early works, convolutional neural fabrics @cite_22 and interlinked CNNs @cite_55 , lack careful design on when to start low-resolution parallel streams and how and when to exchange information across parallel streams, and do not use batch normalization and residual connections, thus not showing satisfactory performance. | {
"cite_N": [
"@cite_55",
"@cite_22"
],
"mid": [
"2295744361",
"2964081403"
],
"abstract": [
"Face parsing is a basic task in face image analysis. It amounts to labeling each pixel with appropriate facial parts such as eyes and nose. In the paper, we present a interlinked convolutional neural network iCNN for solving this problem in an end-to-end fashion. It consists of multiple convolutional neural networks CNNs taking input in different scales. A special interlinking layer is designed to allow the CNNs to exchange information, enabling them to integrate local and contextual information efficiently. The hallmark of iCNN is the extensive use of downsampling and upsampling in the interlinking layers, while traditional CNNs usually uses downsampling only. A two-stage pipeline is proposed for face parsing and both stages use iCNN. The first stage localizes facial parts in the size-reduced image and the second stage labels the pixels in the identified facial parts in the original image. On a benchmark dataset we have obtained better results than the state-of-the-art methods.",
"Despite the success of CNNs, selecting the optimal architecture for a given task remains an open problem. Instead of aiming to select a single optimal architecture, we propose a \" fabric \" that embeds an exponentially large number of architectures. The fabric consists of a 3D trellis that connects response maps at different layers, scales, and channels with a sparse homogeneous local connectivity pattern. The only hyper-parameters of a fabric are the number of channels and layers. While individual architectures can be recovered as paths, the fabric can in addition ensemble all embedded architectures together, sharing their weights where their paths overlap. Parameters can be learned using standard methods based on back-propagation, at a cost that scales linearly in the fabric size. We present benchmark results competitive with the state of the art for image classification on MNIST and CIFAR10, and for semantic segmentation on the Part Labels dataset."
]
} |
1904.04514 | 2940262938 | High-resolution representation learning plays an essential role in many vision problems, e.g., pose estimation and semantic segmentation. The high-resolution network (HRNet) SunXLW19 , recently developed for human pose estimation, maintains high-resolution representations through the whole process by connecting high-to-low resolution convolutions in and produces strong high-resolution representations by repeatedly conducting fusions across parallel convolutions. In this paper, we conduct a further study on high-resolution representations by introducing a simple yet effective modification and apply it to a wide range of vision tasks. We augment the high-resolution representation by aggregating the (upsampled) representations from all the parallel convolutions rather than only the representation from the high-resolution convolution as done in SunXLW19 . This simple modification leads to stronger representations, evidenced by superior results. We show top results in semantic segmentation on Cityscapes, LIP, and PASCAL Context, and facial landmark detection on AFLW, COFW, @math W, and WFLW. In addition, we build a multi-level representation from the high-resolution representation and apply it to the Faster R-CNN object detection framework and the extended frameworks. The proposed approach achieves superior results to existing single-model networks on COCO object detection. The code and models have been publicly available at this https URL . | GridNet @cite_89 is like a combination of multiple U-Nets and includes two symmetric information exchange stages: the first stage only passes information from high-resolution to low-resolution, and the second stage only passes information from low-resolution to high-resolution. This limits its segmentation quality. | {
"cite_N": [
"@cite_89"
],
"mid": [
"2963741402"
],
"abstract": [
"This paper presents GridNet, a new Convolutional Neural Network (CNN) architecture for semantic image segmentation (full scene labelling). Classical neural networks are implemented as one stream from the input to the output with subsampling operators applied in the stream in order to reduce the feature maps size and to increase the receptive field for the final prediction. However, for semantic image segmentation, where the task consists in providing a semantic class to each pixel of an image, feature maps reduction is harmful because it leads to a resolution loss in the output prediction. To tackle this problem, our GridNet follows a grid pattern allowing multiple interconnected streams to work at different resolutions. We show that our network generalizes many well known networks such as conv-deconv, residual or U-Net networks. GridNet is trained from scratch and achieves competitive results on the Cityscapes dataset."
]
} |
1904.04587 | 2938292067 | We propose a new randomized coordinate descent method for unconstrained minimization of convex functions that are smooth with respect to a given real symmetric positive semidefinite curvature matrix. At each iteration, this method works with a random coordinate subset selected with probability proportional to the determinant of the corresponding principal submatrix of the curvature matrix (volume sampling). When the size of the subset equals one, the method reduces to the well-known randomized coordinate descent which samples coordinates with probabilities proportional to the coordinate Lipschitz constants. We establish the convergence rates both for convex and strongly convex functions. Our theoretical results show that by increasing the size of the subset from one number to another, it is possible to accelerate the method up to the factor which depends on the spectral gap between the corresponding largest eigenvalues of the curvature matrix. Several numerical experiments confirm our theoretical conclusions. | One of the most influential papers on coordinate descent is the work @cite_9 by Nesterov, where he proposed a coordinate gradient method (which we will refer to as RCD) with a special rule for selecting coordinates. In RCD, each coordinate is sampled with probability proportional to the corresponding coordinate Lipschitz constant. Nesterov then derived the complexity bound for RCD and showed that it can be even better than that of the standard gradient method. Our method generalizes RCD in the sense that it coincides with it in the special case when the number of coordinates selected at each iteration equals one. | {
"cite_N": [
"@cite_9"
],
"mid": [
"2095984592"
],
"abstract": [
"In this paper we propose new methods for solving huge-scale optimization problems. For problems of this size, even the simplest full-dimensional vector operations are very expensive. Hence, we propose to apply an optimization technique based on random partial update of decision variables. For these methods, we prove the global estimates for the rate of convergence. Surprisingly enough, for certain classes of objective functions, our results are better than the standard worst-case bounds for deterministic algorithms. We present constrained and unconstrained versions of the method, and its accelerated variant. Our numerical test confirms a high efficiency of this technique on problems of very big size."
]
} |
1904.04587 | 2938292067 | We propose a new randomized coordinate descent method for unconstrained minimization of convex functions that are smooth with respect to a given real symmetric positive semidefinite curvature matrix. At each iteration, this method works with a random coordinate subset selected with probability proportional to the determinant of the corresponding principal submatrix of the curvature matrix (volume sampling). When the size of the subset equals one, the method reduces to the well-known randomized coordinate descent which samples coordinates with probabilities proportional to the coordinate Lipschitz constants. We establish the convergence rates both for convex and strongly convex functions. Our theoretical results show that by increasing the size of the subset from one number to another, it is possible to accelerate the method up to the factor which depends on the spectral gap between the corresponding largest eigenvalues of the curvature matrix. Several numerical experiments confirm our theoretical conclusions. | @cite_22 , the authors propose three different randomized methods for unconstrained minimization of a smooth function. Their Method 1 is exactly the same method as ours with the only difference that they consider an sampling for selecting coordinates. As a result, the convergence rate of their method is expressed in terms of the minimal eigenvalue of the expectation of some matrix, and it is not clear how this quantity depends on the number of coordinates used at each iteration. Although the authors consider a particular example of a @math matrix for which their method should be very efficient, they do not establish any general results. In this regard, our work can be considered a further development of @cite_22 , where we establish more interpretable complexity results for a particular kind of coordinate sampling, namely volume sampling. In addition to that, we also consider both the convex and strongly convex cases, while the authors of @cite_22 only consider the latter. | {
"cite_N": [
"@cite_22"
],
"mid": [
"134812657"
],
"abstract": [
"We propose a new algorithm for minimizing regularized empirical loss: Stochastic Dual Newton Ascent (SDNA). Our method is dual in nature: in each iteration we update a random subset of the dual variables. However, unlike existing methods such as stochastic dual coordinate ascent, SDNA is capable of utilizing all local curvature information contained in the examples, which leads to striking improvements in both theory and practice -- sometimes by orders of magnitude. In the special case when an L2-regularizer is used in the primal, the dual problem is a concave quadratic maximization problem plus a separable term. In this regime, SDNA in each step solves a proximal subproblem involving a random principal submatrix of the Hessian of the quadratic function; whence the name of the method."
]
} |
1904.04587 | 2938292067 | We propose a new randomized coordinate descent method for unconstrained minimization of convex functions that are smooth with respect to a given real symmetric positive semidefinite curvature matrix. At each iteration, this method works with a random coordinate subset selected with probability proportional to the determinant of the corresponding principal submatrix of the curvature matrix (volume sampling). When the size of the subset equals one, the method reduces to the well-known randomized coordinate descent which samples coordinates with probabilities proportional to the coordinate Lipschitz constants. We establish the convergence rates both for convex and strongly convex functions. Our theoretical results show that by increasing the size of the subset from one number to another, it is possible to accelerate the method up to the factor which depends on the spectral gap between the corresponding largest eigenvalues of the curvature matrix. Several numerical experiments confirm our theoretical conclusions. | Another closely related work to this one is @cite_17 , where the authors propose a new randomized optimization method for minimizing quadratic functions given @math eigenvectors corresponding to the @math smallest eigenvalues of the Hessian. Although the complexity estimates for this method look similar to ours, there are several key differences. First, the results in @cite_17 show that the increase in the number of coordinates from @math to @math in their method leads to the acceleration rate that depends on the spectral gap between the @math st and @math nd eigenvalues. The corresponding acceleration rate of our method depends, on the contrary, on the spectral gap between the @math st and @math nd eigenvalues. Second, their method is not, strictly speaking, a coordinate descent algorithm since it uses eigenvectors as search directions instead of the coordinate directions. Finally, it is also less practical than ours. For example, even in the simplest non-trivial regime @math , it requires an eigenvector corresponding to the smallest eigenvalue; the complexity of obtaining such a vector is in general @math . In contrast, the simplest non-trivial choice for our method is @math which can be implemented in time @math (or even smaller for sparse problems). | {
"cite_N": [
"@cite_17"
],
"mid": [
"2963651662"
],
"abstract": [
"The state-of-the-art methods for solving optimization problems in big dimensions are variants of randomized coordinate descent (RCD). In this paper we introduce a fundamentally new type of acceleration strategy for RCD based on the augmentation of the set of coordinate directions by a few spectral or conjugate directions. As we increase the number of extra directions to be sampled from, the rate of the method improves, and interpolates between the linear rate of RCD and a linear rate independent of the condition number. We develop and analyze also inexact variants of these methods where the spectral and conjugate directions are allowed to be approximate only. We motivate the above development by proving several negative results which highlight the limitations of RCD with importance sampling."
]
} |
1904.04587 | 2938292067 | We propose a new randomized coordinate descent method for unconstrained minimization of convex functions that are smooth with respect to a given real symmetric positive semidefinite curvature matrix. At each iteration, this method works with a random coordinate subset selected with probability proportional to the determinant of the corresponding principal submatrix of the curvature matrix (volume sampling). When the size of the subset equals one, the method reduces to the well-known randomized coordinate descent which samples coordinates with probabilities proportional to the coordinate Lipschitz constants. We establish the convergence rates both for convex and strongly convex functions. Our theoretical results show that by increasing the size of the subset from one number to another, it is possible to accelerate the method up to the factor which depends on the spectral gap between the corresponding largest eigenvalues of the curvature matrix. Several numerical experiments confirm our theoretical conclusions. | Finally, we should mention that volume sampling is not a novel concept and has already been known in the literature for some time. To our knowledge, it was first proposed in @cite_19 for the problem of matrix approximation. Later on the same authors developed several efficient exact and approximate methods for doing volume sampling @cite_13 based on the standard linear algebra algorithms. Some other polynomial-time sampling methods and their connection to the theory of Markov chains were considered in @cite_16 . Recently volume sampling has also been applied to the problem of linear regression @cite_18 @cite_4 . | {
"cite_N": [
"@cite_18",
"@cite_4",
"@cite_19",
"@cite_16",
"@cite_13"
],
"mid": [
"2963214110",
"2807660787",
"2103318769",
"2728264528",
"2089135543"
],
"abstract": [
"We study the following basic machine learning task: Given a fixed set of input points in ℝd for a linear regression problem, we wish to predict a hidden response value for each of the points. We can only afford to attain the responses for a small subset of the points that are then used to construct linear predictions for all points in the dataset. The performance of the predictions is evaluated by the total square loss on all responses (the attained as well as the remaining hidden ones). We show that a good approximate solution to this least squares problem can be obtained from just dimension d many responses by using a joint sampling technique called volume sampling. Moreover, the least squares solution obtained for the volume sampled subproblem is an unbiased estimator of optimal solution based on all n responses. This unbiasedness is a desirable property that is not shared by other common subset selection techniques. Motivated by these basic properties, we develop a theoretical framework for studying volume sampling, resulting in a number of new matrix expectation equalities and statistical guarantees which are of importance not only to least squares regression but also to numerical linear algebra in general. Our methods also lead to a regularized variant of volume sampling, and we propose the first efficient algorithm for volume sampling which makes this technique a practical tool in the machine learning toolbox. Finally, we provide experimental evidence which confirms our theoretical findings.",
"Suppose an n x d design matrix in a linear regression problem is given, but the response for each point is hidden unless explicitly requested. The goal is to sample only a small number k << n of the responses, and then produce a weight vector whose sum of squares loss over all points is at most 1+epsilon times the minimum. When k is very small (e.g., k=d), jointly sampling diverse subsets of points is crucial. One such method called \"volume sampling\" has a unique and desirable property that the weight vector it produces is an unbiased estimate of the optimum. It is therefore natural to ask if this method offers the optimal unbiased estimate in terms of the number of responses k needed to achieve a 1+epsilon loss approximation. Surprisingly we show that volume sampling can have poor behavior when we require a very accurate approximation -- indeed worse than some i.i.d. sampling techniques whose estimates are biased, such as leverage score sampling. We then develop a new rescaled variant of volume sampling that produces an unbiased estimate which avoids this bad behavior and has at least as good a tail bound as leverage score sampling: sample size k=O(d log d + d epsilon) suffices to guarantee total loss at most 1+epsilon times the minimum with high probability. Thus, we improve on the best previously known sample size for an unbiased estimator, k=O(d^2 epsilon). Our rescaling procedure leads to a new efficient algorithm for volume sampling which is based on a \"determinantal rejection sampling\" technique with potentially broader applications to determinantal point processes. Other contributions include introducing the combinatorics needed for rescaled volume sampling and developing tail bounds for sums of dependent random matrices which arise in the process.",
"[17] proved that a small sample of rows of a given matrix A contains a low-rank approximation D that minimizes ||A - D||F to within small additive error, and the sampling can be done efficiently using just two passes over the matrix [12]. In this paper, we generalize this result in two ways. First, we prove that the additive error drops exponentially by iterating the sampling in an adaptive manner. Using this result, we give a pass-efficient algorithm for computing low-rank approximation with reduced additive error. Our second result is that using a natural distribution on subsets of rows (called volume sampling), there exists a subset of k rows whose span contains a factor (k + 1) relative approximation and a subset of k + k(k + 1) e rows whose span contains a 1+e relative approximation. The existence of such a small certificate for multiplicative low-rank approximation leads to a PTAS for the following projective clustering problem: Given a set of points P in Rd, and integers k, j, find a set of j subspaces F 1 , . . ., F j , each of dimension at most k, that minimize Σ p∈P min i d(p, F i )2.",
"We study dual volume sampling, a method for selecting k columns from an n*m short and wide matrix (n <= k <= m) such that the probability of selection is proportional to the volume spanned by the rows of the induced submatrix. This method was proposed by Avron and Boutsidis (2013), who showed it to be a promising method for column subset selection and its multiple applications. However, its wider adoption has been hampered by the lack of polynomial time sampling algorithms. We remove this hindrance by developing an exact (randomized) polynomial time sampling algorithm as well as its derandomization. Thereafter, we study dual volume sampling via the theory of real stable polynomials and prove that its distribution satisfies the “Strong Rayleigh” property. This result has numerous consequences, including a provably fast-mixing Markov chain sampler that makes dual volume sampling much more attractive to practitioners. This sampler is closely related to classical algorithms for popular experimental design methods that are to date lacking theoretical analysis but are known to empirically work well.",
"We give efficient algorithms for volume sampling, i.e., for picking @math -subsets of the rows of any given matrix with probabilities proportional to the squared volumes of the simplices defined by them and the origin (or the squared volumes of the parallelepipeds defined by these subsets of rows). In other words, we can efficiently sample @math -subsets of @math with probabilities proportional to the corresponding @math by @math principal minors of any given @math by @math positive semi definite matrix. This solves an open problem from the monograph on spectral algorithms by Kannan and Vempala (see Section @math of KV , also implicit in BDM, DRVW ). Our first algorithm for volume sampling @math -subsets of rows from an @math -by- @math matrix runs in @math arithmetic operations (where @math is the exponent of matrix multiplication) and a second variant of it for @math -approximate volume sampling runs in @math arithmetic operations, which is almost linear in the size of the input (i.e., the number of entries) for small @math . Our efficient volume sampling algorithms imply the following results for low-rank matrix approximation: (1) Given @math , in @math arithmetic operations we can find @math of its rows such that projecting onto their span gives a @math -approximation to the matrix of rank @math closest to @math under the Frobenius norm. This improves the @math -approximation of Boutsidis, Drineas and Mahoney BDM and matches the lower bound shown in DRVW . The method of conditional expectations gives a algorithm with the same complexity. The running time can be improved to @math at the cost of losing an extra @math in the approximation factor. (2) The same rows and projection as in the previous point give a @math -approximation to the matrix of rank @math closest to @math under the spectral norm. In this paper, we show an almost matching lower bound of @math , even for @math ."
]
} |
1904.04706 | 2939969660 | We study the problem of safety verification of direct perception neural networks, which take camera images as inputs and produce high-level features for autonomous vehicles to make control decisions. Formal verification of direct perception neural networks is extremely challenging, as it is difficult to formulate the specification that requires characterizing input conditions, while the number of neurons in such a network can reach millions. We approach the specification problem by learning an input property characterizer which carefully extends a direct perception neural network at close-to-output layers, and address the scalability problem by only analyzing networks starting from shared neurons without losing soundness. The presented workflow is used to understand a direct perception neural network (developed by Audi) which computes the next waypoint and orientation for autonomous vehicles to follow. | Formal verification of neural networks has drawn huge attention with many results available @cite_10 @cite_17 @cite_18 @cite_12 @cite_0 @cite_14 @cite_9 @cite_15 @cite_11 @cite_6 @cite_2 @cite_7 . Although specifications used in formal verification of neural networks are discussed in recent reviews @cite_19 @cite_13 , the specification problem over images is not addressed, so research results largely use inherent properties of a neural network such as local robustness (as output invariance) or output ranges where one does not need to characterize properties for desired inputs. Not being able to properly characterizing input conditions (one possibility is to simply consider every input to be bounded by @math ) makes it difficult for formal static analysis to achieve any useful results on deep perception networks, regardless of the type of abstraction domain being used (box, octagon, or zonotope). Lastly, our work is motivated by @cite_16 which trains additional features apart from a standard neural network. The feature detector is commonly created by extending the network from close-to-output layers. | {
"cite_N": [
"@cite_13",
"@cite_18",
"@cite_14",
"@cite_11",
"@cite_7",
"@cite_9",
"@cite_6",
"@cite_0",
"@cite_19",
"@cite_2",
"@cite_15",
"@cite_16",
"@cite_10",
"@cite_12",
"@cite_17"
],
"mid": [
"2893132739",
"2963600714",
"2963673089",
"2801079363",
"2917325587",
"2794609696",
"2800473699",
"2721006554",
"2904076963",
"2963440492",
"2791251367",
"2150295085",
"2165073069",
"2963054787",
"2594877703"
],
"abstract": [
"The increasing use of deep neural networks in a variety of applications, including some safety-critical ones, has brought renewed interest in the topic of verification of neural networks. However, verification is most meaningful when performed with high-quality formal specifications. In this paper, we survey the landscape of formal specification for deep neural networks, and discuss the opportunities and challenges for formal methods for this domain.",
"The deployment of Artificial Neural Networks (ANNs) in safety-critical applications poses a number of new verification and certification challenges. In particular, for ANN-enabled self-driving vehicles it is important to establish properties about the resilience of ANNs to noisy or even maliciously manipulated sensory input. We are addressing these challenges by defining resilience properties of ANN-based classifiers as the maximum amount of input or sensor perturbation which is still tolerated. This problem of computing maximum perturbation bounds for ANNs is then reduced to solving mixed integer optimization problems (MIP). A number of MIP encoding heuristics are developed for drastically reducing MIP-solver runtimes, and using parallelization of MIP-solvers results in an almost linear speed-up in the number (up to a certain limit) of computing cores in our experiments. We demonstrate the effectiveness and scalability of our approach by means of computing maximum resilience bounds for a number of ANN benchmark sets ranging from typical image recognition scenarios to the autonomous maneuvering of robots.",
"",
"The success of Deep Learning and its potential use in many safety-critical applications has motivated research on formal verification of Neural Network (NN) models. Despite the reputation of learned NN models to behave as black boxes and the theoretical hardness of proving their properties, researchers have been successful in verifying some classes of models by exploiting their piecewise linear structure and taking insights from formal methods such as Satisifiability Modulo Theory. These methods are however still far from scaling to realistic neural networks. To facilitate progress on this crucial area, we make two key contributions. First, we present a unified framework that encompasses previous methods. This analysis results in the identification of new methods that combine the strengths of multiple existing approaches, accomplishing a speedup of two orders of magnitude compared to the previous state of the art. Second, we propose a new data set of benchmarks which includes a collection of previously released testcases. We use the benchmark to provide the first experimental comparison of existing algorithms and identify the factors impacting the hardness of verification problems.",
"Deep neural networks (DNNs) have been shown lack of robustness for the vulnerability of their classification to small perturbations on the inputs. This has led to safety concerns of applying DNNs to safety-critical domains. Several verification approaches have been developed to automatically prove or disprove safety properties of DNNs. However, these approaches suffer from either the scalability problem, i.e., only small DNNs can be handled, or the precision problem, i.e., the obtained bounds are loose. This paper improves on a recent proposal of analyzing DNNs through the classic abstract interpretation technique, by a novel symbolic propagation technique. More specifically, the values of neurons are represented symbolically and propagated forwardly from the input layer to the output layer, on top of abstract domains. We show that our approach can achieve significantly higher precision and thus can prove more properties than using only abstract domains. Moreover, we show that the bounds derived from our approach on the hidden neurons, when applied to a state-of-the-art SMT based verification tool, can improve its performance. We implement our approach into a software tool and validate it over a few DNNs trained on benchmark datasets such as MNIST, etc.",
"We present AI2, the first sound and scalable analyzer for deep neural networks. Based on overapproximation, AI2 can automatically prove safety properties (e.g., robustness) of realistic neural networks (e.g., convolutional neural networks). The key insight behind AI2 is to phrase reasoning about safety and robustness of neural networks in terms of classic abstract interpretation, enabling us to leverage decades of advances in that area. Concretely, we introduce abstract transformers that capture the behavior of fully connected and convolutional neural network layers with rectified linear unit activations (ReLU), as well as max pooling layers. This allows us to handle real-world neural networks, which are often built out of those types of layers. We present a complete implementation of AI2 together with an extensive evaluation on 20 neural networks. Our results demonstrate that: (i) AI2 is precise enough to prove useful specifications (e.g., robustness), (ii) AI2 can be used to certify the effectiveness of state-of-the-art defenses for neural networks, (iii) AI2 is significantly faster than existing analyzers based on symbolic analysis, which often take hours to verify simple fully connected networks, and (iv) AI2 can handle deep convolutional networks, which are beyond the reach of existing methods.",
"Verifying correctness of deep neural networks (DNNs) is challenging. We study a generic reachability problem for feed-forward DNNs which, for a given set of inputs to the network and a Lipschitz-continuous function over its outputs, computes the lower and upper bound on the function values. Because the network and the function are Lipschitz continuous, all values in the interval between the lower and upper bound are reachable. We show how to obtain the safety verification problem, the output range analysis problem and a robustness measure by instantiating the reachability problem. We present a novel algorithm based on adaptive nested optimisation to solve the reachability problem. The technique has been implemented and evaluated on a range of DNNs, demonstrating its efficiency, scalability and ability to handle a broader class of networks than state-of-the-art verification approaches.",
"We study the reachability problem for systems implemented as feed-forward neural networks whose activation function is implemented via ReLU functions. We draw a correspondence between establishing whether some arbitrary output can ever be outputed by a neural system and linear problems characterising a neural system of interest. We present a methodology to solve cases of practical interest by means of a state-of-the-art linear programs solver. We evaluate the technique presented by discussing the experimental results obtained by analysing reachability properties for a number of benchmarks in the literature.",
"In the past few years, significant progress has been made on deep neural networks (DNNs) in achieving human-level intelligence on several long-standing tasks. With broader deployment of DNNs on various applications, the concerns on its safety and trustworthiness have been raised, particularly after the fatal incidents of self-driving cars. Research to address these concerns is very active, with many papers released in the past few years. This survey paper is to conduct a review of the current research efforts on making DNNs safe and trustworthy, by focusing on four aspects, i.e., verification, testing, adversarial attack and defence, and interpretability. In total, we surveyed 178 papers, most of which were published in the most recent two years, i.e., 2017 and 2018.",
"",
"Given a neural network (NN) and a set of possible inputs to the network described by polyhedral constraints, we aim to compute a safe over-approximation of the set of possible output values. This operation is a fundamental primitive enabling the formal analysis of neural networks that are extensively used in a variety of machine learning tasks such as perception and control of autonomous systems. Increasingly, they are deployed in high-assurance applications, leading to a compelling use case for formal verification approaches. In this paper, we present an efficient range estimation algorithm that iterates between an expensive global combinatorial search using mixed-integer linear programming problems, and a relatively inexpensive local optimization that repeatedly seeks a local optimum of the function represented by the NN. We implement our approach and compare it with Reluplex, a recently proposed solver for deep neural networks. We demonstrate applications of our approach to computing flowpipes for neural network-based feedback controllers. We show that the use of local search in conjunction with mixed-integer linear programming solvers effectively reduces the combinatorial search over possible combinations of active neurons in the network by pruning away suboptimal nodes.",
"We consider the problem of zero-shot learning, where the goal is to learn a classifier f : X → Y that must predict novel values of Y that were omitted from the training set. To achieve this, we define the notion of a semantic output code classifier (SOC) which utilizes a knowledge base of semantic properties of Y to extrapolate to novel classes. We provide a formalism for this type of classifier and study its theoretical properties in a PAC framework, showing conditions under which the classifier can accurately predict novel classes. As a case study, we build a SOC classifier for a neural decoding task and show that it can often predict words that people are thinking about from functional magnetic resonance images (fMRI) of their neural activity, even without training examples for those words.",
"A key problem in the adoption of artificial neural networks in safety-related applications is that misbehaviors can be hardly ruled out with traditional analytical or probabilistic techniques In this paper we focus on specific networks known as Multi-Layer Perceptrons (MLPs), and we propose a solution to verify their safety using abstractions to Boolean combinations of linear arithmetic constraints We show that our abstractions are consistent, i.e., whenever the abstract MLP is declared to be safe, the same holds for the concrete one Spurious counterexamples, on the other hand, trigger refinements and can be leveraged to automate the correction of misbehaviors We describe an implementation of our approach based on the HySAT solver, detailing the abstraction-refinement process and the automated correction strategy Finally, we present experimental results confirming the feasibility of our approach on a realistic case study.",
"We present an approach for the verification of feed-forward neural networks in which all nodes have a piece-wise linear activation function. Such networks are often used in deep learning and have been shown to be hard to verify for modern satisfiability modulo theory (SMT) and integer linear programming (ILP) solvers.",
"Deep neural networks have emerged as a widely used and effective means for tackling complex, real-world problems. However, a major obstacle in applying them to safety-critical systems is the great difficulty in providing formal guarantees about their behavior. We present a novel, scalable, and efficient technique for verifying properties of deep neural networks (or providing counter-examples). The technique is based on the simplex method, extended to handle the non-convex Rectified Linear Unit (ReLU) activation function, which is a crucial ingredient in many modern neural networks. The verification procedure tackles neural networks as a whole, without making any simplifying assumptions. We evaluated our technique on a prototype deep neural network implementation of the next-generation airborne collision avoidance system for unmanned aircraft (ACAS Xu). Results show that our technique can successfully prove properties of networks that are an order of magnitude larger than the largest networks verified using existing methods."
]
} |
1904.04776 | 2949857013 | Accurate prediction of others' trajectories is essential for autonomous driving. Trajectory prediction is challenging because it requires reasoning about agents' past movements, social interactions among varying numbers and kinds of agents, constraints from the scene context, and the stochasticity of human behavior. Our approach models these interactions and constraints jointly within a novel Multi-Agent Tensor Fusion (MATF) network. Specifically, the model encodes multiple agents' past trajectories and the scene context into a Multi-Agent Tensor, then applies convolutional fusion to capture multiagent interactions while retaining the spatial structure of agents and the scene context. The model decodes recurrently to multiple agents' future trajectories, using adversarial loss to learn stochastic predictions. Experiments on both highway driving and pedestrian crowd datasets show that the model achieves state-of-the-art prediction accuracy. | Agent-centric NN-based approaches integrate information from multiple agents by applying aggregation functions on multiple agents' feature vectors output from recurrent units. Social LSTM @cite_17 runs max pooling over state vectors of nearby agents within a predefined distance range, but does not model social interaction with far-away agents. Social GAN @cite_19 contributes a new pooling mechanism over all the agents involved in a scene globally, and by using adversarial training to learn a stochastic, generative model of human behavior @cite_7 . Although these kinds of max pooling aggregation functions handle varying numbers of agents well, permutation invariant functions may discard information when input agents lose their uniqueness @cite_3 . In contrast, Social Attention @cite_21 and Sophie @cite_3 address the heterogeneity of social interaction among different agents by attention mechanisms @cite_20 @cite_11 , and spatial-temporal graphs @cite_13 . Attention mechanisms encode which other agents are most important to focus on when predicting the trajectory of a given agent. However, attention-based approaches are very sensitive to the number of agents included --- predicting @math agents has @math computational complexity. In contrast, our approach captures multiagent interactions while maintaining @math computational complexity. | {
"cite_N": [
"@cite_11",
"@cite_7",
"@cite_21",
"@cite_3",
"@cite_19",
"@cite_13",
"@cite_20",
"@cite_17"
],
"mid": [
"",
"2099471712",
"2766836212",
"2805102305",
"2963001155",
"",
"2964308564",
"2424778531"
],
"abstract": [
"",
"We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a mistake. This framework corresponds to a minimax two-player game. In the space of arbitrary functions G and D, a unique solution exists, with G recovering the training data distribution and D equal to ½ everywhere. In the case where G and D are defined by multilayer perceptrons, the entire system can be trained with backpropagation. There is no need for any Markov chains or unrolled approximate inference networks during either training or generation of samples. Experiments demonstrate the potential of the framework through qualitative and quantitative evaluation of the generated samples.",
"Robots that navigate through human crowds need to be able to plan safe, efficient, and human predictable trajectories. This is a particularly challenging problem as it requires the robot to predict future human trajectories within a crowd where everyone implicitly cooperates with each other to avoid collisions. Previous approaches to human trajectory prediction have modeled the interactions between humans as a function of proximity. However, that is not necessarily true as some people in our immediate vicinity moving in the same direction might not be as important as other people that are further away, but that might collide with us in the future. In this work, we propose Social Attention, a novel trajectory prediction model that captures the relative importance of each person when navigating in the crowd, irrespective of their proximity. We demonstrate the performance of our method against a state-of-the-art approach on two publicly available crowd datasets and analyze the trained attention model to gain a better understanding of which surrounding agents humans attend to, when navigating in a crowd.",
"This paper addresses the problem of path prediction for multiple interacting agents in a scene, which is a crucial step for many autonomous platforms such as self-driving cars and social robots. We present ; an interpretable framework based on Generative Adversarial Network (GAN), which leverages two sources of information, the path history of all the agents in a scene, and the scene context information, using images of the scene. To predict a future path for an agent, both physical and social information must be leveraged. Previous work has not been successful to jointly model physical and social interactions. Our approach blends a social attention mechanism with a physical attention that helps the model to learn where to look in a large scene and extract the most salient parts of the image relevant to the path. Whereas, the social attention component aggregates information across the different agent interactions and extracts the most important trajectory information from the surrounding neighbors. SoPhie also takes advantage of GAN to generates more realistic samples and to capture the uncertain nature of the future paths by modeling its distribution. All these mechanisms enable our approach to predict socially and physically plausible paths for the agents and to achieve state-of-the-art performance on several different trajectory forecasting benchmarks.",
"Understanding human motion behavior is critical for autonomous moving platforms (like self-driving cars and social robots) if they are to navigate human-centric environments. This is challenging because human motion is inherently multimodal: given a history of human motion paths, there are many socially plausible ways that people could move in the future. We tackle this problem by combining tools from sequence prediction and generative adversarial networks: a recurrent sequence-to-sequence model observes motion histories and predicts future behavior, using a novel pooling mechanism to aggregate information across people. We predict socially plausible futures by training adversarially against a recurrent discriminator, and encourage diverse predictions with a novel variety loss. Through experiments on several datasets we demonstrate that our approach outperforms prior work in terms of accuracy, variety, collision avoidance, and computational complexity.",
"",
"Abstract: Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.",
"Pedestrians follow different trajectories to avoid obstacles and accommodate fellow pedestrians. Any autonomous vehicle navigating such a scene should be able to foresee the future positions of pedestrians and accordingly adjust its path to avoid collisions. This problem of trajectory prediction can be viewed as a sequence generation task, where we are interested in predicting the future trajectory of people based on their past positions. Following the recent success of Recurrent Neural Network (RNN) models for sequence prediction tasks, we propose an LSTM model which can learn general human movement and predict their future trajectories. This is in contrast to traditional approaches which use hand-crafted functions such as Social forces. We demonstrate the performance of our method on several public datasets. Our model outperforms state-of-the-art methods on some of these datasets. We also analyze the trajectories predicted by our model to demonstrate the motion behaviour learned by our model."
]
} |
1904.04776 | 2949857013 | Accurate prediction of others' trajectories is essential for autonomous driving. Trajectory prediction is challenging because it requires reasoning about agents' past movements, social interactions among varying numbers and kinds of agents, constraints from the scene context, and the stochasticity of human behavior. Our approach models these interactions and constraints jointly within a novel Multi-Agent Tensor Fusion (MATF) network. Specifically, the model encodes multiple agents' past trajectories and the scene context into a Multi-Agent Tensor, then applies convolutional fusion to capture multiagent interactions while retaining the spatial structure of agents and the scene context. The model decodes recurrently to multiple agents' future trajectories, using adversarial loss to learn stochastic predictions. Experiments on both highway driving and pedestrian crowd datasets show that the model achieves state-of-the-art prediction accuracy. | Many data-driven approaches learn to predict deterministic future trajectories of agents by minimizing reconstruction loss @cite_17 @cite_6 . However, human behavior is inherently stochastic. Recent approaches address this by predicting a distribution over future trajectories by combining Variational Auto-Encoders @cite_28 and Inverse Optimal Control @cite_36 , or with conditional Generative Adversarial Nets @cite_19 @cite_40 . GAIL-GRU @cite_23 uses generative adversarial imitation learning @cite_18 to learn a stochastic policy that reproduces human expert driving behavior. R2P2 @cite_37 proposes a novel cost function to encourage enhancement in both precision and diversity of the learned predictive distribution. Other approaches predict a set of possible trajectories, instead of a single deterministic trajectory, by conditioning on possible maneuver classes @cite_29 @cite_4 . | {
"cite_N": [
"@cite_18",
"@cite_37",
"@cite_4",
"@cite_28",
"@cite_36",
"@cite_29",
"@cite_6",
"@cite_19",
"@cite_40",
"@cite_23",
"@cite_17"
],
"mid": [
"2963277051",
"2894978157",
"",
"",
"2607296803",
"2963906196",
"",
"2963001155",
"",
"2580495915",
"2424778531"
],
"abstract": [
"Consider learning a policy from example expert behavior, without interaction with the expert or access to a reinforcement signal. One approach is to recover the expert's cost function with inverse reinforcement learning, then extract a policy from that cost function with reinforcement learning. This approach is indirect and can be slow. We propose a new general framework for directly extracting a policy from data as if it were obtained by reinforcement learning following inverse reinforcement learning. We show that a certain instantiation of our framework draws an analogy between imitation learning and generative adversarial networks, from which we derive a model-free imitation learning algorithm that obtains significant performance gains over existing model-free methods in imitating complex behaviors in large, high-dimensional environments.",
"We propose a method to forecast a vehicle’s ego-motion as a distribution over spatiotemporal paths, conditioned on features (e.g., from LIDAR and images) embedded in an overhead map. The method learns a policy inducing a distribution over simulated trajectories that is both “diverse” (produces most of the likely paths) and “precise” (mostly produces likely paths). This balance is achieved through minimization of a symmetrized cross-entropy between the distribution and demonstration data. By viewing the simulated-outcome distribution as the pushforward of a simple distribution under a simulation operator, we obtain expressions for the cross-entropy metrics that can be efficiently evaluated and differentiated, enabling stochastic-gradient optimization. We propose concrete policy architectures for this model, discuss our evaluation metrics relative to previously-used degenerate metrics, and demonstrate the superiority of our method relative to state-of-the-art methods in both the Kitti dataset and a similar but novel and larger real-world dataset explicitly designed for the vehicle forecasting domain.",
"",
"",
"We introduce a Deep Stochastic IOC RNN Encoder-decoder framework, DESIRE, for the task of future predictions of multiple interacting agents in dynamic scenes. DESIRE effectively predicts future locations of objects in multiple scenes by 1) accounting for the multi-modal nature of the future prediction (i.e., given the same context, future may vary), 2) foreseeing the potential future outcomes and make a strategic prediction based on that, and 3) reasoning not only from the past motion history, but also from the scene context as well as the interactions among the agents. DESIRE achieves these in a single end-to-end trainable neural network model, while being computationally efficient. The model first obtains a diverse set of hypothetical future prediction samples employing a conditional variational auto-encoder, which are ranked and refined by the following RNN scoring-regression module. Samples are scored by accounting for accumulated future rewards, which enables better long-term strategic decisions similar to IOC frameworks. An RNN scene context fusion module jointly captures past motion histories, the semantic scene context and interactions among multiple agents. A feedback mechanism iterates over the ranking and refinement to further boost the prediction accuracy. We evaluate our model on two publicly available datasets: KITTI and Stanford Drone Dataset. Our experiments show that the proposed model significantly improves the prediction accuracy compared to other baseline methods.",
"Forecasting the motion of surrounding vehicles is a critical ability for an autonomous vehicle deployed in complex traffic. Motion of all vehicles in a scene is governed by the traffic context, i.e., the motion and relative spatial configuration of neighboring vehicles. In this paper we propose an LSTM encoder-decoder model that uses convolutional social pooling as an improvement to social pooling layers for robustly learning interdependencies in vehicle motion. Additionally, our model outputs a multi-modal predictive distribution over future trajectories based on maneuver classes. We evaluate our model using the publicly available NGSIM US-101 and I-80 datasets. Our results show improvement over the state of the art in terms of RMS values of prediction error and negative log-likelihoods of true future trajectories under the model's predictive distribution. We also present a qualitative analysis of the model's predicted distributions for various traffic scenarios.",
"",
"Understanding human motion behavior is critical for autonomous moving platforms (like self-driving cars and social robots) if they are to navigate human-centric environments. This is challenging because human motion is inherently multimodal: given a history of human motion paths, there are many socially plausible ways that people could move in the future. We tackle this problem by combining tools from sequence prediction and generative adversarial networks: a recurrent sequence-to-sequence model observes motion histories and predicts future behavior, using a novel pooling mechanism to aggregate information across people. We predict socially plausible futures by training adversarially against a recurrent discriminator, and encourage diverse predictions with a novel variety loss. Through experiments on several datasets we demonstrate that our approach outperforms prior work in terms of accuracy, variety, collision avoidance, and computational complexity.",
"",
"The ability to accurately predict and simulate human driving behavior is critical for the development of intelligent transportation systems. Traditional modeling methods have employed simple parametric models and behavioral cloning. This paper adopts a method for overcoming the problem of cascading errors inherent in prior approaches, resulting in realistic behavior that is robust to trajectory perturbations. We extend Generative Adversarial Imitation Learning to the training of recurrent policies, and we demonstrate that our model rivals rule-based controllers and maximum likelihood models in realistic highway simulations. Our model both reproduces emergent behavior of human drivers, such as lane change rate, while maintaining realistic control over long time horizons.",
"Pedestrians follow different trajectories to avoid obstacles and accommodate fellow pedestrians. Any autonomous vehicle navigating such a scene should be able to foresee the future positions of pedestrians and accordingly adjust its path to avoid collisions. This problem of trajectory prediction can be viewed as a sequence generation task, where we are interested in predicting the future trajectory of people based on their past positions. Following the recent success of Recurrent Neural Network (RNN) models for sequence prediction tasks, we propose an LSTM model which can learn general human movement and predict their future trajectories. This is in contrast to traditional approaches which use hand-crafted functions such as Social forces. We demonstrate the performance of our method on several public datasets. Our model outperforms state-of-the-art methods on some of these datasets. We also analyze the trajectories predicted by our model to demonstrate the motion behaviour learned by our model."
]
} |
1904.04661 | 2936269784 | In radiologists' routine work, one major task is to read a medical image, e.g., a CT scan, find significant lesions, and describe them in the radiology report. In this paper, we study the lesion description or annotation problem. Given a lesion image, our aim is to predict a comprehensive set of relevant labels, such as the lesion's body part, type, and attributes, which may assist downstream fine-grained diagnosis. To address this task, we first design a deep learning module to extract relevant semantic labels from the radiology reports associated with the lesion images. With the images and text-mined labels, we propose a lesion annotation network (LesaNet) based on a multilabel convolutional neural network (CNN) to learn all labels holistically. Hierarchical relations and mutually exclusive relations between the labels are leveraged to improve the label prediction accuracy. The relations are utilized in a label expansion strategy and a relational hard example mining algorithm. We also attach a simple score propagation layer on LesaNet to enhance recall and explore implicit relation between labels. Multilabel metric learning is combined with classification to enable interpretable prediction. We evaluated LesaNet on the public DeepLesion dataset, which contains over 32K diverse lesion images. Experiments show that LesaNet can precisely annotate the lesions using an ontology of 171 fine-grained labels with an average AUC of 0.9344. | Noisy and incomplete training labels often exist in datasets mined from the web @cite_4 , which is similar to our labels mined from reports. Strategies to handle them include data filtering @cite_23 , noise-robust losses @cite_8 , noise modeling @cite_49 , finding reliable negative samples @cite_13 , and so on. We use a text-mining module to filter noisy positive labels and leverage label relation to find reliable negative labels. Label relations have been exploited by researchers to improve image classification. Novel loss functions were proposed in @cite_44 for labels with hierarchical tree-like structures. In @cite_18 @cite_40 , prediction scores of different labels are propagated between network layers whose structure is designed to capture label relations. We apply label expansion and RHEM strategies to use label relations explicitly, and at the same time employ a score propagation layer to learn them implicitly. | {
"cite_N": [
"@cite_18",
"@cite_4",
"@cite_8",
"@cite_44",
"@cite_40",
"@cite_49",
"@cite_23",
"@cite_13"
],
"mid": [
"2581887665",
"2007972815",
"2962762541",
"2697115435",
"",
"2962749380",
"2951558862",
"22461475"
],
"abstract": [
"One characteristic that sets humans apart from modern learning-based computer vision algorithms is the ability to acquire knowledge about the world and use that knowledge to reason about the visual world. Humans can learn about the characteristics of objects and the relationships that occur between them to learn a large variety of visual concepts, often with few examples. This paper investigates the use of structured prior knowledge in the form of knowledge graphs and shows that using this knowledge improves performance on image classification. We build on recent work on end-to-end learning on graphs, introducing the Graph Search Neural Network as a way of efficiently incorporating large knowledge graphs into a vision classification pipeline. We show in a number of experiments that our method outperforms standard neural network baselines for multi-label classification.",
"This paper introduces a web image dataset created by NUS's Lab for Media Search. The dataset includes: (1) 269,648 images and the associated tags from Flickr, with a total of 5,018 unique tags; (2) six types of low-level features extracted from these images, including 64-D color histogram, 144-D color correlogram, 73-D edge direction histogram, 128-D wavelet texture, 225-D block-wise color moments extracted over 5x5 fixed grid partitions, and 500-D bag of words based on SIFT descriptions; and (3) ground-truth for 81 concepts that can be used for evaluation. Based on this dataset, we highlight characteristics of Web image collections and identify four research issues on web image annotation and retrieval. We also provide the baseline results for web image annotation by learning from the tags using the traditional k-NN algorithm. The benchmark results indicate that it is possible to learn effective models from sufficiently large image dataset to facilitate general image retrieval.",
"Current state-of-the-art deep learning systems for visual object recognition and detection use purely supervised training with regularization such as dropout to avoid overfitting. The performance depends critically on the amount of labeled examples, and in current practice the labels are assumed to be unambiguous and accurate. However, this assumption often does not hold; e.g. in recognition, class labels may be missing; in detection, objects in the image may not be localized; and in general, the labeling may be subjective. In this work we propose a generic way to handle noisy and incomplete labeling by augmenting the prediction objective with a notion of consistency. We consider a prediction consistent if the same prediction is made given similar percepts, where the notion of similarity is between deep network features computed from the input data. In experiments we demonstrate that our approach yields substantial robustness to label noise on several datasets. On MNIST handwritten digits, we show that our model is robust to label corruption. On the Toronto Face Database, we show that our model handles well the case of subjective labels in emotion recognition, achieving state-of-theart results, and can also benefit from unlabeled face images with no modification to our method. On the ILSVRC2014 detection challenge data, we show that our approach extends to very deep networks, high resolution images and structured outputs, and results in improved scalable detection.",
"Neural networks are powerful tools for medical image classification and segmentation. However, existing network structures and training procedures assume that the output classes are mutually exclusive and equally important. Many datasets of medical images do not satisfy these conditions. For example, some skin disease datasets have images labelled as coarse-grained class (such as Benign) in addition to images with fine-grained labels (such as a Benign subclass called Blue Nevus), and conventional neural network can not leverage such additional data for training. Also, in the clinical decision making, some classes (such as skin cancer or Melanoma) often carry more importance than other lesion types. We propose a novel Tree-Loss function for training and fine-tuning a neural network classifier using all available labelled images. The key step is the definition of the class taxonomy tree, which is used to describe the relations between labels. The tree can be also adjusted to reflect the desired importance of each class. These steps can be performed by a domain expert without detailed knowledge of machine learning techniques. The experiments demonstrate the improved performance compared with the conventional approach even without using additional data.",
"",
"When human annotators are given a choice about what to label in an image, they apply their own subjective judgments on what to ignore and what to mention. We refer to these noisy \"human-centric\" annotations as exhibiting human reporting bias. Examples of such annotations include image tags and keywords found on photo sharing sites, or in datasets containing image captions. In this paper, we use these noisy annotations for learning visually correct image classifiers. Such annotations do not use consistent vocabulary, and miss a significant amount of the information present in an image, however, we demonstrate that the noise in these annotations exhibits structure and can be modeled. We propose an algorithm to decouple the human reporting bias from the correct visually grounded labels. Our results are highly interpretable for reporting \"what's in the image\" versus \"what's worth saying.\" We demonstrate the algorithm's efficacy along a variety of metrics and datasets, including MS COCO and Yahoo Flickr 100M.We show significant improvements over traditional algorithms for both image classification and image captioning, doubling the performance of existing methods in some cases.",
"Current approaches for fine-grained recognition do the following: First, recruit experts to annotate a dataset of images, optionally also collecting more structured data in the form of part annotations and bounding boxes. Second, train a model utilizing this data. Toward the goal of solving fine-grained recognition, we introduce an alternative approach, leveraging free, noisy data from the web and simple, generic methods of recognition. This approach has benefits in both performance and scalability. We demonstrate its efficacy on four fine-grained datasets, greatly exceeding existing state of the art without the manual collection of even a single label, and furthermore show first results at scaling to more than 10,000 fine-grained categories. Quantitatively, we achieve top-1 accuracies of 92.3 on CUB-200-2011, 85.4 on Birdsnap, 93.4 on FGVC-Aircraft, and 80.8 on Stanford Dogs without using their annotated training sets. We compare our approach to an active learning approach for expanding fine-grained datasets.",
"In traditional text classification, a classifier is built using labeled training documents of every class. This paper studies a different problem. Given a set P of documents of a particular class (called positive class) and a set U of unlabeled documents that contains documents from class P and also other types of documents (called negative class documents), we want to build a classifier to classify the documents in U into documents from P and documents not from P. The key feature of this problem is that there is no labeled negative document, which makes traditional text classification techniques inapplicable. In this paper, we propose an effective technique to solve the problem. It combines the Rocchio method and the SVM technique for classifier building. Experimental results show that the new method outperforms existing methods significantly."
]
} |
1904.04617 | 2938853100 | In this paper, we study the joint power control and scheduling in uplink massive multiple-input multiple-output (MIMO) systems with random data arrivals. The data is generated at each user according to an individual stochastic process. Using Lyapunov optimization techniques, we develop a dynamic scheduling algorithm (DSA), which decides at each time slot the amount of data to admit to the transmission queues and the transmission rates over the wireless channel. The proposed algorithm achieves nearly optimal performance on the long-term user throughput under various fairness policies. Simulation results show that the DSA can improve the time-average delay performance compared to the state-of-the-art power control schemes developed for Massive MIMO with infinite backlogs. | The basis of stochastic network optimization was presented in @cite_2 , with many examples of its applications to communication and queueing systems. Flow control algorithms in a heterogeneous network using Lyapunov optimization techniques were proposed in @cite_8 . Furthermore, @cite_5 gives an overview of cross-layer control and scheduling algorithms using the drift-plus-penalty framework to achieve stability while optimizing some network performance metric. MIMO downlink scheduling using the flow control algorithm was studied in @cite_9 , where the backlog at the BS is assumed to be infinite. Due to the difficulty of solving the weighted sum rate maximization problem, an on-off scheduling policy was considered as an approximation of the optimal rate allocation problem. A dynamic scheduling scheme using Lyapunov techniques was developed in @cite_3 for millimeter wave-enabled Massive MIMO systems, after some simplifying assumptions on the rate expressions. Note that @cite_12 proposes an effective algorithm to solve the weighted sum rate maximization problem for Massive MIMO systems with i.i.d. Rayleigh fading channels, which facilitate the application of Lyapunov optimization in Massive MIMO with random data traffic. | {
"cite_N": [
"@cite_8",
"@cite_9",
"@cite_3",
"@cite_2",
"@cite_5",
"@cite_12"
],
"mid": [
"2120344179",
"2149819557",
"2612103911",
"2137152139",
"2138993731",
"2561146563"
],
"abstract": [
"We consider optimal control for general networks with both wireless and wireline components and time varying channels. A dynamic strategy is developed to support all traffic whenever possible, and to make optimally fair decisions about which data to serve when inputs exceed network capacity. The strategy is decoupled into separate algorithms for flow control, routing, and resource allocation, and allows each user to make decisions independent of the actions of others. The combined strategy is shown to yield data rates that are arbitrarily close to the optimal operating point achieved when all network controllers are coordinated and have perfect knowledge of future events. The cost of approaching this fair operating point is an end-to-end delay increase for data that is served by the network.",
"Downlink scheduling schemes are well-known and widely investigated under the assumption that the channel state is perfectly known to the scheduler. In the multiuser MIMO (broadcast) case, downlink scheduling in the presence of non-perfect channel state information (CSI) is only scantly treated. In this paper we provide a general framework that addresses the problem systematically. Also, we illuminate the key role played by the channel state prediction error: our scheme treats in a fundamentally different way users with small channel prediction error (\"predictable\" users) and users with large channel prediction error (\"non-predictable\" users), and can be interpreted as a near-optimal opportunistic time-sharing strategy between MIMO downlink beamforming to predictable users and space-time coding to non-predictable users. Our results, based on a realistic MIMO channel model used in 3GPP standardization, show that the proposed algorithms can significantly outperform a conventional \"mismatched\" scheduling scheme that treats the available CSI as if it was perfect.",
"Ultra-reliability and low latency are two key components in 5G networks. In this letter, we investigate the problem of ultra-reliable and low-latency communication in millimeter wave-enabled massive multiple-input multiple-output networks. The problem is cast as a network utility maximization subject to probabilistic latency and reliability constraints. To solve this problem, we resort to the Lyapunov technique, whereby a utility-delay control approach is proposed, which adapts to channel variations and queue dynamics. Numerical results demonstrate that our proposed approach ensures reliable communication with a guaranteed probability of 99.99 , and reduces latency by 28.41 and 77.11 as compared to baselines with and without probabilistic latency constraints, respectively.",
"This text presents a modern theory of analysis, control, and optimization for dynamic networks. Mathematical techniques of Lyapunov drift and Lyapunov optimization are developed and shown to enable constrained optimization of time averages in general stochastic systems. The focus is on communication and queueing systems, including wireless networks with time-varying channels, mobility, and randomly arriving traffic. A simple drift-plus-penalty framework is used to optimize time averages such as throughput, throughput-utility, power, and distortion. Explicit performance-delay tradeoffs are provided to illustrate the cost of approaching optimality. This theory is also applicable to problems in operations research and economics, where energy-efficient and profit-maximizing decisions must be made without knowing the future. Topics in the text include the following: - Queue stability theory - Backpressure, max-weight, and virtual queue methods - Primal-dual methods for non-convex stochastic utility maximization - Universal scheduling theory for arbitrary sample paths - Approximate and randomized scheduling theory - Optimization of renewal systems and Markov decision systems Detailed examples and numerous problem set questions are provided to reinforce the main concepts. Table of Contents: Introduction Introduction to Queues Dynamic Scheduling Example Optimizing Time Averages Optimizing Functions of Time Averages Approximate Scheduling Optimization of Renewal Systems Conclusions",
"Information flow in a telecommunication network is accomplished through the interaction of mechanisms at various design layers with the end goal of supporting the information exchange needs of the applications. In wireless networks in particular, the different layers interact in a nontrivial manner in order to support information transfer. In this text we will present abstract models that capture the cross-layer interaction from the physical to transport layer in wireless network architectures including cellular, ad-hoc and sensor networks as well as hybrid wireless-wireline. The model allows for arbitrary network topologies as well as traffic forwarding modes, including datagrams and virtual circuits. Furthermore the time varying nature of a wireless network, due either to fading channels or to changing connectivity due to mobility, is adequately captured in our model to allow for state dependent network control policies. Quantitative performance measures that capture the quality of service requirements in these systems depending on the supported applications are discussed, including throughput maximization, energy consumption minimization, rate utility function maximization as well as general performance functionals. Cross-layer control algorithms with optimal or suboptimal performance with respect to the above measures are presented and analyzed. A detailed exposition of the related analysis and design techniques is provided.",
"This paper considers the jointly optimal pilot and data power allocation in single-cell uplink massive multiple-input-multiple-output systems. Using the spectral efficiency (SE) as performance metric and setting a total energy budget per coherence interval, the power control is formulated as optimization problems for two different objective functions: the weighted minimum SE among the users and the weighted sum SE. A closed form solution for the optimal length of the pilot sequence is derived. The optimal power control policy for the former problem is found by solving a simple equation with a single variable. Utilizing the special structure arising from imperfect channel estimation, a convex reformulation is found to solve the latter problem to global optimality in polynomial time. The gain of the optimal joint power control is theoretically justified, and is proved to be large in the low-SNR regime. Simulation results also show the advantage of optimizing the power control over both pilot and data power, as compared to the cases of using full power and of only optimizing the data powers as done in previous work."
]
} |
1904.04689 | 2938050891 | Recognising actions in videos relies on labelled supervision during training, typically the start and end times of each action instance. This supervision is not only subjective, but also expensive to acquire. Weak video-level supervision has been successfully exploited for recognition in untrimmed videos, however it is challenged when the number of different actions in training videos increases. We propose a method that is supervised by single timestamps located around each action instance, in untrimmed videos. We replace expensive action bounds with sampling distributions initialised from these timestamps. We then use the classifier's response to iteratively update the sampling distributions. We demonstrate that these distributions converge to the location and extent of discriminative action segments. We evaluate our method on three datasets for fine-grained recognition, with increasing number of different actions per video, and show that single timestamps offer a reasonable compromise between recognition performance and labelling effort, performing comparably to full temporal supervision. Our update method improves top-1 test accuracy by up to 5.4 . across the evaluated datasets. | We review recent works using weak temporal labels for action recognition and localisation. For a review of works that use strong supervision, we refer the reader to @cite_15 . We divide the section into works using video-level, transcript and point-level supervision. | {
"cite_N": [
"@cite_15"
],
"mid": [
"2963218601"
],
"abstract": [
"Understanding human actions in visual data is tied to advances in complementary research areas including object recognition, human dynamics, domain adaptation and semantic segmentation. Over the last decade, human action analysis evolved from earlier schemes that are often limited to controlled environments to nowadays advanced solutions that can learn from millions of videos and apply to almost all daily activities. Given the broad range of applications from video surveillance to humancomputer interaction, scientific milestones in action recognition are achieved more rapidly, eventually leading to the demise of what used to be good in a short time. This motivated us to provide a comprehensive review of the notable steps taken towards recognizing human actions. To this end, we start our discussion with the pioneering methods that use handcrafted representations, and then, navigate into the realm of deep learning based approaches. We aim to remain objective throughout this survey, touching upon encouraging improvements as well as inevitable fallbacks, in the hope of raising fresh questions and motivating new research directions for the reader. We provide a detailed review of the work on human action recognition over the past decade.We refer to actions as meaningful human motions.Including Hand-crafted representations methods, we review the impact of Deep-nets on action recognition.We follow a systematic taxonomy to highlight the essence of both Hand-crafted and Deep-net solutions.We present a comparison of methods at their algorithmic level and performance."
]
} |
1904.04689 | 2938050891 | Recognising actions in videos relies on labelled supervision during training, typically the start and end times of each action instance. This supervision is not only subjective, but also expensive to acquire. Weak video-level supervision has been successfully exploited for recognition in untrimmed videos, however it is challenged when the number of different actions in training videos increases. We propose a method that is supervised by single timestamps located around each action instance, in untrimmed videos. We replace expensive action bounds with sampling distributions initialised from these timestamps. We then use the classifier's response to iteratively update the sampling distributions. We demonstrate that these distributions converge to the location and extent of discriminative action segments. We evaluate our method on three datasets for fine-grained recognition, with increasing number of different actions per video, and show that single timestamps offer a reasonable compromise between recognition performance and labelling effort, performing comparably to full temporal supervision. Our update method improves top-1 test accuracy by up to 5.4 . across the evaluated datasets. | provides the weakest cue, signalling only the presence or absence of an action in an untrimmed video, discarding any temporal ordering. When only a few different actions are present in an untrimmed video, video-level supervision can prove sufficient to learn the actions even from long videos, as recently shown in @cite_25 @cite_4 @cite_30 @cite_16 @cite_34 . In these works, the authors use such supervision to train a model for action classification and localisation, achieving results often comparable to those obtained with strongly supervised approaches. However, all these works evaluate their approach on the THUMOS 14 @cite_3 and Activity Net @cite_31 datasets, which contain mainly one class per training video. In this work, we show that as the number of different actions per training video increases, video-level labels do not offer sufficient supervision. | {
"cite_N": [
"@cite_30",
"@cite_4",
"@cite_34",
"@cite_3",
"@cite_31",
"@cite_16",
"@cite_25"
],
"mid": [
"2963045696",
"2962709777",
"2884293275",
"",
"1927052826",
"2895240652",
"2604113307"
],
"abstract": [
"We propose ‘Hide-and-Seek’, a weakly-supervised framework that aims to improve object localization in images and action localization in videos. Most existing weakly-supervised methods localize only the most discriminative parts of an object rather than all relevant parts, which leads to suboptimal performance. Our key idea is to hide patches in a training image randomly, forcing the network to seek other relevant parts when the most discriminative part is hidden. Our approach only needs to modify the input image and can work with any network designed for object localization. During testing, we do not need to hide any patches. Our Hide-and-Seek approach obtains superior performance compared to previous methods for weakly-supervised object localization on the ILSVRC dataset. We also demonstrate that our framework can be easily extended to weakly-supervised action localization.",
"We propose a weakly supervised temporal action localization algorithm on untrimmed videos using convolutional neural networks. Our algorithm learns from video-level class labels and predicts temporal intervals of human actions with no requirement of temporal localization annotations. We design our network to identify a sparse subset of key segments associated with target actions in a video using an attention module and fuse the key segments through adaptive temporal pooling. Our loss function is comprised of two terms that minimize the video-level action classification error and enforce the sparsity of the segment selection. At inference time, we extract and score temporal proposals using temporal class activations and class-agnostic attentions to estimate the time intervals that correspond to target actions. The proposed algorithm attains state-of-the-art results on the THUMOS14 dataset and outstanding performance on ActivityNet1.3 even with its weak supervision.",
"Most activity localization methods in the literature suffer from the burden of frame-wise annotation requirement. Learning from weak labels may be a potential solution towards reducing such manual labeling effort. Recent years have witnessed a substantial influx of tagged videos on the Internet, which can serve as a rich source of weakly-supervised training data. Specifically, the correlations between videos with similar tags can be utilized to temporally localize the activities. Towards this goal, we present W-TALC, a Weakly-supervised Temporal Activity Localization and Classification framework using only video-level labels. The proposed network can be divided into two sub-networks, namely the Two-Stream based feature extractor network and a weakly-supervised module, which we learn by optimizing two complimentary loss functions. Qualitative and quantitative results on two challenging datasets - Thumos14 and ActivityNet1.2, demonstrate that the proposed method is able to detect activities at a fine granularity and achieve better performance than current state-of-the-art methods.",
"",
"In spite of many dataset efforts for human action recognition, current computer vision algorithms are still severely limited in terms of the variability and complexity of the actions that they can recognize. This is in part due to the simplicity of current benchmarks, which mostly focus on simple actions and movements occurring on manually trimmed videos. In this paper we introduce ActivityNet, a new large-scale video benchmark for human activity understanding. Our benchmark aims at covering a wide range of complex human activities that are of interest to people in their daily living. In its current version, ActivityNet provides samples from 203 activity classes with an average of 137 untrimmed videos per class and 1.41 activity instances per video, for a total of 849 video hours. We illustrate three scenarios in which ActivityNet can be used to compare algorithms for human activity understanding: untrimmed video classification, trimmed activity classification and activity detection.",
"Temporal Action Localization (TAL) in untrimmed video is important for many applications. But it is very expensive to annotate the segment-level ground truth (action class and temporal boundary). This raises the interest of addressing TAL with weak supervision, namely only video-level annotations are available during training). However, the state-of-the-art weakly-supervised TAL methods only focus on generating good Class Activation Sequence (CAS) over time but conduct simple thresholding on CAS to localize actions. In this paper, we first develop a novel weakly-supervised TAL framework called AutoLoc to directly predict the temporal boundary of each action instance. We propose a novel Outer-Inner-Contrastive (OIC) loss to automatically discover the needed segment-level supervision for training such a boundary predictor. Our method achieves dramatically improved performance: under the IoU threshold 0.5, our method improves mAP on THUMOS’14 from 13.7 to 21.2 and mAP on ActivityNet from 7.4 to 27.3 . It is also very encouraging to see that our weakly-supervised method achieves comparable results with some fully-supervised methods.",
"Current action recognition methods heavily rely on trimmed videos for model training. However, it is expensive and time-consuming to acquire a large-scale trimmed video dataset. This paper presents a new weakly supervised architecture, called UntrimmedNet, which is able to directly learn action recognition models from untrimmed videos without the requirement of temporal annotations of action instances. Our UntrimmedNet couples two important components, the classification module and the selection module, to learn the action models and reason about the temporal duration of action instances, respectively. These two components are implemented with feed-forward networks, and UntrimmedNet is therefore an end-to-end trainable architecture. We exploit the learned models for action recognition (WSR) and detection (WSD) on the untrimmed video datasets of THUMOS14 and ActivityNet. Although our UntrimmedNet only employs weak supervision, our method achieves performance superior or comparable to that of those strongly supervised approaches on these two datasets."
]
} |
1904.04689 | 2938050891 | Recognising actions in videos relies on labelled supervision during training, typically the start and end times of each action instance. This supervision is not only subjective, but also expensive to acquire. Weak video-level supervision has been successfully exploited for recognition in untrimmed videos, however it is challenged when the number of different actions in training videos increases. We propose a method that is supervised by single timestamps located around each action instance, in untrimmed videos. We replace expensive action bounds with sampling distributions initialised from these timestamps. We then use the classifier's response to iteratively update the sampling distributions. We demonstrate that these distributions converge to the location and extent of discriminative action segments. We evaluate our method on three datasets for fine-grained recognition, with increasing number of different actions per video, and show that single timestamps offer a reasonable compromise between recognition performance and labelling effort, performing comparably to full temporal supervision. Our update method improves top-1 test accuracy by up to 5.4 . across the evaluated datasets. | offers an ordered list of action labels in untrimmed videos, without any temporal annotations @cite_11 @cite_26 @cite_0 @cite_2 @cite_5 @cite_29 @cite_14 @cite_7 . Some works @cite_7 @cite_5 @cite_2 assume the transcript includes knowledge of background', specifying whether the actions occur in succession or with gaps. @cite_7 , uniform sampling of the video is followed by iterative refinement of the action boundaries. The refinement uses the pairwise comparison of softmax scores for class labels around each boundary, along with linear interpolation. This iterative boundary refinement strategy is conceptually similar to ours. However, the approach in @cite_7 assumes no gaps are allowed between neighbouring actions. This requires knowledge of background labels in order for the method to operate. | {
"cite_N": [
"@cite_26",
"@cite_14",
"@cite_7",
"@cite_29",
"@cite_0",
"@cite_2",
"@cite_5",
"@cite_11"
],
"mid": [
"262578090",
"2962916463",
"2964311439",
"2600081845",
"2477205648",
"2463824207",
"2530494944",
"2108710284"
],
"abstract": [
"Suppose that we are given a set of videos, along with natural language descriptions in the form of multiple sentences (e.g., manual annotations, movie scripts, sport summaries etc.), and that these sentences appear in the same temporal order as their visual counterparts. We propose in this paper a method for aligning the two modalities, i.e., automatically providing a time (frame) stamp for every sentence. Given vectorial features for both video and text, this can be cast as a temporal assignment problem, with an implicit linear mapping between the two feature modalities. We formulate this problem as an integer quadratic program, and solve its continuous convex relaxation using an efficient conditional gradient algorithm. Several rounding procedures are proposed to construct the final integer solution. After demonstrating significant improvements over the state of the art on the related task of aligning video with symbolic labels [7], we evaluate our method on a challenging dataset of videos with associated textual descriptions [37], and explore bag-of-words and continuous representations for text.",
"Video learning is an important task in computer vision and has experienced increasing interest over the recent years. Since even a small amount of videos easily comprises several million frames, methods that do not rely on a frame-level annotation are of special importance. In this work, we propose a novel learning algorithm with a Viterbi-based loss that allows for online and incremental learning of weakly annotated video data. We moreover show that explicit context and length modeling leads to huge improvements in video segmentation and labeling tasks and include these models into our framework. On several action segmentation benchmarks, we obtain an improvement of up to 10 compared to current state-of-the-art methods.",
"In this work, we address the task of weakly-supervised human action segmentation in long, untrimmed videos. Recent methods have relied on expensive learning models, such as Recurrent Neural Networks (RNN) and Hidden Markov Models (HMM). However, these methods suffer from expensive computational cost, thus are unable to be deployed in large scale. To overcome the limitations, the keys to our design are efficiency and scalability. We propose a novel action modeling framework, which consists of a new temporal convolutional network, named Temporal Convolutional Feature Pyramid Network (TCFPN), for predicting frame-wise action labels, and a novel training strategy for weakly-supervised sequence modeling, named Iterative Soft Boundary Assignment (ISBA), to align action sequences and update the network in an iterative fashion. The proposed framework is evaluated on two benchmark datasets, Breakfast and Hollywood Extended, with four different evaluation metrics. Extensive experimental results show that our methods achieve competitive or superior performance to state-of-the-art methods.",
"We present an approach for weakly supervised learning of human actions. Given a set of videos and an ordered list of the occurring actions, the goal is to infer start and end frames of the related action classes within the video and to train the respective action classifiers without any need for hand labeled frame boundaries. To address this task, we propose a combination of a discriminative representation of subactions, modeled by a recurrent neural network, and a coarse probabilistic model to allow for a temporal alignment and inference over long sequences. While this system alone already generates good results, we show that the performance can be further improved by approximating the number of subactions to the characteristics of the different action classes. To this end, we adapt the number of subaction classes by iterating realignment and reestimation during training. The proposed system is evaluated on two benchmark datasets, the Breakfast and the Hollywood extended dataset, showing a competitive performance on various weak learning tasks such as temporal action segmentation and action alignment.",
"We propose a weakly-supervised framework for action labeling in video, where only the order of occurring actions is required during training time. The key challenge is that the per-frame alignments between the input (video) and label (action) sequences are unknown during training. We address this by introducing the Extended Connectionist Temporal Classification (ECTC) framework to efficiently evaluate all possible alignments via dynamic programming and explicitly enforce their consistency with frame-to-frame visual similarities. This protects the model from distractions of visually inconsistent or degenerated alignments without the need of temporal supervision. We further extend our framework to the semi-supervised case when a few frames are sparsely annotated in a video. With less than 1 of labeled frames per video, our method is able to outperform existing semi-supervised approaches and achieve comparable performance to that of fully supervised approaches.",
"While current approaches to action recognition on presegmented video clips already achieve high accuracies, temporal action detection is still far from comparably good results. Automatically locating and classifying the relevant action segments in videos of varying lengths proves to be a challenging task. We propose a novel method for temporal action detection including statistical length and language modeling to represent temporal and contextual structure. Our approach aims at globally optimizing the joint probability of three components, a length and language model and a discriminative action model, without making intermediate decisions. The problem of finding the most likely action sequence and the corresponding segment boundaries in an exponentially large search space is addressed by dynamic programming. We provide an extensive evaluation of each model component on Thumos 14, a large action detection dataset, and report state-of-the-art results on three datasets.",
"Abstract We present an approach for weakly supervised learning of human actions from video transcriptions. Our system is based on the idea that, given a sequence of input data and a transcript, i.e. a list of the order the actions occur in the video, it is possible to infer the actions within the video stream and to learn the related action models without the need for any frame-based annotation. Starting from the transcript information at hand, we split the given data sequences uniformly based on the number of expected actions. We then learn action models for each class by maximizing the probability that the training video sequences are generated by the action models given the sequence order as defined by the transcripts. The learned model can be used to temporally segment an unseen video with or without transcript. Additionally, the inferred segments can be used as a starting point to train high-level fully supervised models. We evaluate our approach on four distinct activity datasets, namely Hollywood Extended, MPII Cooking, Breakfast and CRIM13. It shows that the proposed system is able to align the scripted actions with the video data, that the learned models localize and classify actions in the datasets, and that they outperform any current state-of-the-art approach for aligning transcripts with video data.",
"We are given a set of video clips, each one annotated with an ordered list of actions, such as “walk” then “sit” then “answer phone” extracted from, for example, the associated text script. We seek to temporally localize the individual actions in each clip as well as to learn a discriminative classifier for each action. We formulate the problem as a weakly supervised temporal assignment with ordering constraints. Each video clip is divided into small time intervals and each time interval of each video clip is assigned one action label, while respecting the order in which the action labels appear in the given annotations. We show that the action label assignment can be determined together with learning a classifier for each action in a discriminative manner. We evaluate the proposed model on a new and challenging dataset of 937 video clips with a total of 787720 frames containing sequences of 16 different actions from 69 Hollywood movies."
]
} |
1904.04689 | 2938050891 | Recognising actions in videos relies on labelled supervision during training, typically the start and end times of each action instance. This supervision is not only subjective, but also expensive to acquire. Weak video-level supervision has been successfully exploited for recognition in untrimmed videos, however it is challenged when the number of different actions in training videos increases. We propose a method that is supervised by single timestamps located around each action instance, in untrimmed videos. We replace expensive action bounds with sampling distributions initialised from these timestamps. We then use the classifier's response to iteratively update the sampling distributions. We demonstrate that these distributions converge to the location and extent of discriminative action segments. We evaluate our method on three datasets for fine-grained recognition, with increasing number of different actions per video, and show that single timestamps offer a reasonable compromise between recognition performance and labelling effort, performing comparably to full temporal supervision. Our update method improves top-1 test accuracy by up to 5.4 . across the evaluated datasets. | refers to using a single pixel or a single frame as a form of supervision. This was attempted for semantic segmentation, by annotating single points in static images @cite_23 and subsequently used for videos @cite_18 @cite_20 . @cite_18 a single pixel is used to annotate the action, in a subset of frames, both spatially and temporally. When combining this weak supervision with action proposals, the authors show that it is possible to achieve comparable results to those obtained with full and much more expensive per-frame bounding boxes. More recently, several forms of weak supervision, including single temporal points, are evaluated in @cite_20 for the task of spatio-temporal action localisation. This work uses an off-the-shelf human detector to extract human tracks from the videos, integrating these with the various annotations in a unified framework based on discriminative clustering. | {
"cite_N": [
"@cite_18",
"@cite_20",
"@cite_23"
],
"mid": [
"2398000642",
"2950064629",
"611457968"
],
"abstract": [
"We strive for spatio-temporal localization of actions in videos. The state-of-the-art relies on action proposals at test time and selects the best one with a classifier trained on carefully annotated box annotations. Annotating action boxes in video is cumbersome, tedious, and error prone. Rather than annotating boxes, we propose to annotate actions in video with points on a sparse subset of frames only. We introduce an overlap measure between action proposals and points and incorporate them all into the objective of a non-convex Multiple Instance Learning optimization. Experimental evaluation on the UCF Sports and UCF 101 datasets shows that (i) spatio-temporal proposals can be used to train classifiers while retaining the localization performance, (ii) point annotations yield results comparable to box annotations while being significantly faster to annotate, (iii) with a minimum amount of supervision our approach is competitive to the state-of-the-art. Finally, we introduce spatio-temporal action annotations on the train and test videos of Hollywood2, resulting in Hollywood2Tubes, available at http: tinyurl.com hollywood2tubes.",
"Spatio-temporal action detection in videos is typically addressed in a fully-supervised setup with manual annotation of training videos required at every frame. Since such annotation is extremely tedious and prohibits scalability, there is a clear need to minimize the amount of manual supervision. In this work we propose a unifying framework that can handle and combine varying types of less-demanding weak supervision. Our model is based on discriminative clustering and integrates different types of supervision as constraints on the optimization. We investigate applications of such a model to training setups with alternative supervisory signals ranging from video-level class labels over temporal points or sparse action bounding boxes to the full per-frame annotation of action bounding boxes. Experiments on the challenging UCF101-24 and DALY datasets demonstrate competitive performance of our method at a fraction of supervision used by previous methods. The flexibility of our model enables joint learning from data with different levels of annotation. Experimental results demonstrate a significant gain by adding a few fully supervised examples to otherwise weakly labeled videos.",
"The semantic image segmentation task presents a trade-off between test time accuracy and training time annotation cost. Detailed per-pixel annotations enable training accurate models but are very time-consuming to obtain; image-level class labels are an order of magnitude cheaper but result in less accurate models. We take a natural step from image-level annotation towards stronger supervision: we ask annotators to point to an object if one exists. We incorporate this point supervision along with a novel objectness potential in the training loss function of a CNN model. Experimental results on the PASCAL VOC 2012 benchmark reveal that the combined effect of point-level supervision and objectness potential yields an improvement of (12.9 , ) mIOU over image-level supervision. Further, we demonstrate that models trained with point-level supervision are more accurate than models trained with image-level, squiggle-level or full supervision given a fixed annotation budget."
]
} |
1904.04689 | 2938050891 | Recognising actions in videos relies on labelled supervision during training, typically the start and end times of each action instance. This supervision is not only subjective, but also expensive to acquire. Weak video-level supervision has been successfully exploited for recognition in untrimmed videos, however it is challenged when the number of different actions in training videos increases. We propose a method that is supervised by single timestamps located around each action instance, in untrimmed videos. We replace expensive action bounds with sampling distributions initialised from these timestamps. We then use the classifier's response to iteratively update the sampling distributions. We demonstrate that these distributions converge to the location and extent of discriminative action segments. We evaluate our method on three datasets for fine-grained recognition, with increasing number of different actions per video, and show that single timestamps offer a reasonable compromise between recognition performance and labelling effort, performing comparably to full temporal supervision. Our update method improves top-1 test accuracy by up to 5.4 . across the evaluated datasets. | In this work, we also use a single temporal point per action for fine-grained recognition in videos. However, unlike the works above @cite_18 @cite_20 which consider the given annotations correct, we actively refine the temporal scope of the given supervision, under the assumption that the given annotated points may be misaligned with the actions and thus lead to incorrect supervision. We show this to effectively converge, when tested on three datasets with varying complexity, in the number of different actions in untrimmed training videos. We detail our method next. | {
"cite_N": [
"@cite_18",
"@cite_20"
],
"mid": [
"2398000642",
"2950064629"
],
"abstract": [
"We strive for spatio-temporal localization of actions in videos. The state-of-the-art relies on action proposals at test time and selects the best one with a classifier trained on carefully annotated box annotations. Annotating action boxes in video is cumbersome, tedious, and error prone. Rather than annotating boxes, we propose to annotate actions in video with points on a sparse subset of frames only. We introduce an overlap measure between action proposals and points and incorporate them all into the objective of a non-convex Multiple Instance Learning optimization. Experimental evaluation on the UCF Sports and UCF 101 datasets shows that (i) spatio-temporal proposals can be used to train classifiers while retaining the localization performance, (ii) point annotations yield results comparable to box annotations while being significantly faster to annotate, (iii) with a minimum amount of supervision our approach is competitive to the state-of-the-art. Finally, we introduce spatio-temporal action annotations on the train and test videos of Hollywood2, resulting in Hollywood2Tubes, available at http: tinyurl.com hollywood2tubes.",
"Spatio-temporal action detection in videos is typically addressed in a fully-supervised setup with manual annotation of training videos required at every frame. Since such annotation is extremely tedious and prohibits scalability, there is a clear need to minimize the amount of manual supervision. In this work we propose a unifying framework that can handle and combine varying types of less-demanding weak supervision. Our model is based on discriminative clustering and integrates different types of supervision as constraints on the optimization. We investigate applications of such a model to training setups with alternative supervisory signals ranging from video-level class labels over temporal points or sparse action bounding boxes to the full per-frame annotation of action bounding boxes. Experiments on the challenging UCF101-24 and DALY datasets demonstrate competitive performance of our method at a fraction of supervision used by previous methods. The flexibility of our model enables joint learning from data with different levels of annotation. Experimental results demonstrate a significant gain by adding a few fully supervised examples to otherwise weakly labeled videos."
]
} |
1904.04794 | 2938495304 | We address the problem of cross-modal information retrieval in the domain of remote sensing. In particular, we are interested in two application scenarios: i) cross-modal retrieval between panchromatic (PAN) and multi-spectral imagery, and ii) multi-label image retrieval between very high resolution (VHR) images and speech based label annotations. Notice that these multi-modal retrieval scenarios are more challenging than the traditional uni-modal retrieval approaches given the inherent differences in distributions between the modalities. However, with the growing availability of multi-source remote sensing data and the scarcity of enough semantic annotations, the task of multi-modal retrieval has recently become extremely important. In this regard, we propose a novel deep neural network based architecture which is considered to learn a discriminative shared feature space for all the input modalities, suitable for semantically coherent information retrieval. Extensive experiments are carried out on the benchmark large-scale PAN - multi-spectral DSRSID dataset and the multi-label UC-Merced dataset. Together with the Merced dataset, we generate a corpus of speech signals corresponding to the labels. Superior performance with respect to the current state-of-the-art is observed in all the cases. | The image retrieval frameworks can either be content-based or text-based, depending upon the nature of the query data. The applications of CBIR in RS mainly gained its motion after the Landsat thematic mapper multi-spectral satellite data were made accessible. Subsequently, several techniques came up which focus on training of the retrieval models through the notion of active learning by incorporating the user's knowledge during training iterations @cite_13 . As opposed to the real-valued feature representations, several techniques consider the binary features for quicker response during retrieval. In this regard, @cite_12 proposes a hashing-based approximate nearest neighbor search for a fast and scalable RS image retrieval. Visual databases targeted by CBIR applications have been extended for multi-label image retrieval @cite_7 @cite_6 . | {
"cite_N": [
"@cite_13",
"@cite_7",
"@cite_12",
"@cite_6"
],
"mid": [
"2171586273",
"2766938848",
"2249017129",
""
],
"abstract": [
"User-defined classes in large generalist image databases are often composed of several groups of images and span very different scales in the space of low-level visual descriptors. The interactive retrieval of such image classes is then very difficult. To addess his challenge, we propose and evaluate here two general mprovements of SVM-based relevance feedback methods. First, to optimize the transfer of information between the user and the system, we focus on the criterion employed by the system for selecting the images presented to the user at every feedback round. We put forward a new active learning selection criterion that minimizes redundancy between the candidate images shown to the user. Second, for image classes having very different scales, we find that a high sensitivity of the SVM to the scale of the data brings about a low retrieval performance. We then argue that insensitivity to scale is desirable in this context and we show how to obtain it by the use of specific kernel functions. The experimental evaluation of both ranking and classification performance on several image databases confirms the effectiveness of our selection criterion and of the use of kernels that reduce the sensitivity of SVMs to the scale of the data",
"Conventional supervised content-based remote sensing (RS) image retrieval systems require a large number of already annotated images to train a classifier for obtaining high retrieval accuracy. Most systems assume that each training image is annotated by a single label associated to the most significant semantic content of the image. However, this assumption does not fit well with the complexity of RS images, where an image might have multiple land-cover classes (i.e., multilabels). Moreover, annotating images with multilabels is costly and time consuming. To address these issues, in this paper, we introduce a semisupervised graph-theoretic method in the framework of multilabel RS image retrieval problems. The proposed method is based on four main steps. The first step segments each image in the archive and extracts the features of each region. The second step constructs an image neighborhood graph and uses a correlated label propagation algorithm to automatically assign a set of labels to each image in the archive by exploiting only a small number of training images annotated with multilabels. The third step associates class labels with image regions by a novel region labeling strategy, whereas the final step retrieves the images similar to a given query image by a subgraph matching strategy. Experiments carried out on an archive of aerial images show the effectiveness of the proposed method when compared with the state-of-the-art RS content-based image retrieval methods.",
"Large-scale remote sensing (RS) image search and retrieval have recently attracted great attention, due to the rapid evolution of satellite systems, that results in a sharp growing of image archives. An exhaustive search through linear scan from such archives is time demanding and not scalable in operational applications. To overcome such a problem, this paper introduces hashing-based approximate nearest neighbor search for fast and accurate image search and retrieval in large RS data archives. The hashing aims at mapping high-dimensional image feature vectors into compact binary hash codes, which are indexed into a hash table that enables real-time search and accurate retrieval. Such binary hash codes can also significantly reduce the amount of memory required for storing the RS images in the auxiliary archives. In particular, in this paper, we introduce in RS two kernel-based nonlinear hashing methods. The first hashing method defines hash functions in the kernel space by using only unlabeled images, while the second method leverages on the semantic similarity extracted by annotated images to describe much distinctive hash functions in the kernel space. The effectiveness of considered hashing methods is analyzed in terms of RS image retrieval accuracy and retrieval time. Experiments carried out on an archive of aerial images point out that the presented hashing methods are much faster, while keeping a similar (or even higher) retrieval accuracy, than those typically used in RS, which exploit an exact nearest neighbor search.",
""
]
} |
1904.04794 | 2938495304 | We address the problem of cross-modal information retrieval in the domain of remote sensing. In particular, we are interested in two application scenarios: i) cross-modal retrieval between panchromatic (PAN) and multi-spectral imagery, and ii) multi-label image retrieval between very high resolution (VHR) images and speech based label annotations. Notice that these multi-modal retrieval scenarios are more challenging than the traditional uni-modal retrieval approaches given the inherent differences in distributions between the modalities. However, with the growing availability of multi-source remote sensing data and the scarcity of enough semantic annotations, the task of multi-modal retrieval has recently become extremely important. In this regard, we propose a novel deep neural network based architecture which is considered to learn a discriminative shared feature space for all the input modalities, suitable for semantically coherent information retrieval. Extensive experiments are carried out on the benchmark large-scale PAN - multi-spectral DSRSID dataset and the multi-label UC-Merced dataset. Together with the Merced dataset, we generate a corpus of speech signals corresponding to the labels. Superior performance with respect to the current state-of-the-art is observed in all the cases. | As already mentioned, cross-modal retrieval has mainly received its attention in the image to text or text to image settings @cite_2 @cite_3 @cite_8 @cite_10 . There have also been works on color image for depth perception for improved scene synthesis @cite_17 . However, there exists not much prior research regarding cross-modal retrieval in RS given the scarcity of multi-source images databases suitable for the retrieval task. In this respect, @cite_11 very recently has proposed a large-scale dual-source remote sensing image dataset (DSRSID) which comprises of panchromatic and multi-spectral image pairs acquired from the Gaofen-1 satellite. Another attempt for cross-modal data retrieval in RS is by @cite_15 , which introduces a deep cross-modal retrieval framework for image and audio data. To the best of our knowledge, these are the only two reported works in RS in cross-modal retrieval. However, a major limitation to these works is that both of them are modality specific and not generic, unlike ours. | {
"cite_N": [
"@cite_11",
"@cite_8",
"@cite_3",
"@cite_2",
"@cite_15",
"@cite_10",
"@cite_17"
],
"mid": [
"2808376087",
"2203543769",
"2425857666",
"2740755807",
"2898008292",
"",
"2171740948"
],
"abstract": [
"Due to the urgent demand for remote sensing big data analysis, large-scale remote sensing image retrieval (LSRSIR) attracts increasing attention from researchers. Generally, LSRSIR can be divided into two categories as follows: uni-source LSRSIR (US-LSRSIR) and cross-source LSRSIR (CS-LSRSIR). More specifically, US-LSRSIR means the inquiry remote sensing image and images in the searching data set come from the same remote sensing data source, whereas CS-LSRSIR is designed to retrieve remote sensing images with a similar content to the inquiry remote sensing image that are from a different remote sensing data source. In the literature, US-LSRSIR has been widely exploited, but CS-LSRSIR is rarely discussed. In practical situations, remote sensing images from different kinds of remote sensing data sources are continually increasing, so there is a great motivation to exploit CS-LSRSIR. Therefore, this paper focuses on CS-LSRSIR. To cope with CS-LSRSIR, this paper proposes source-invariant deep hashing convolutional neural networks (SIDHCNNs), which can be optimized in an end-to-end manner using a series of well-designed optimization constraints. To quantitatively evaluate the proposed SIDHCNNs, we construct a dual-source remote sensing image data set that contains eight typical land-cover categories and 10 000 dual samples in each category. Extensive experiments show that the proposed SIDHCNNs can yield substantial improvements over several baselines involving the most recent techniques.",
"Due to its low storage cost and fast query speed, hashing has been widely adopted for similarity search in multimedia data. In particular, more and more attentions have been payed to multimodal hashing for search in multimedia data with multiple modalities, such as images with tags. Typically, supervised information of semantic labels is also available for the data points in many real applications. Hence, many supervised multimodal hashing (SMH) methods have been proposed to utilize such semantic labels to further improve the search accuracy. However, the training time complexity of most existing SMH methods is too high, which makes them unscalable to large-scale datasets. In this paper, a novel SMH method, called semantic correlation maximization (SCM), is proposed to seamlessly integrate semantic labels into the hashing learning procedure for large-scale data modeling. Experimental results on two real-world datasets show that SCM can significantly outperform the state-of-the-art SMH methods, in terms of both accuracy and scalability.",
"Hashing has attracted a great deal of research in recent years due to its effectiveness for the retrieval and indexing of large-scale high-dimensional multimedia data. In this paper, we propose a novel ranking-based hashing framework that maps data from different modalities into a common Hamming space where the cross-modal similarity can be measured using Hamming distance. Unlike existing cross-modal hashing algorithms where the learned hash functions are binary space partitioning functions, such as the sign and threshold function, the proposed hashing scheme takes advantage of a new class of hash functions closely related to rank correlation measures which are known to be scale-invariant, numerically stable, and highly nonlinear. Specifically, we jointly learn two groups of linear subspaces, one for each modality, so that features’ ranking orders in different linear subspaces maximally preserve the cross-modal similarities. We show that the ranking-based hash function has a natural probabilistic approximation which transforms the original highly discontinuous optimization problem into one that can be efficiently solved using simple gradient descent algorithms. The proposed hashing framework is also flexible in the sense that the optimization procedures are not tied up to any specific form of loss function, which is typical for existing cross-modal hashing methods, but rather we can flexibly accommodate different loss functions with minimal changes to the learning steps. We demonstrate through extensive experiments on four widely-used real-world multimodal datasets that the proposed cross-modal hashing method can achieve competitive performance against several state-of-the-arts with only moderate training and testing time.",
"Due to availability of large amounts of multimedia data, cross-modal matching is gaining increasing importance. Hashing based techniques provide an attractive solution to this problem when the data size is large. Different scenarios of cross-modal matching are possible, for example, data from the different modalities can be associated with a single label or multiple labels, and in addition may or may not have one-to-one correspondence. Most of the existing approaches have been developed for the case where there is one-to-one correspondence between the data of the two modalities. In this paper, we propose a simple, yet effective generalized hashing framework which can work for all the different scenarios, while preserving the semantic distance between the data points. The approach first learns the optimum hash codes for the two modalities simultaneously, so as to preserve the semantic similarity between the data points, and then learns the hash functions to map from the features to the hash codes. Extensive experiments on single label dataset like Wiki and multi-label datasets like NUS-WIDE, Pascal and LabelMe under all the different scenarios and comparisons with the state-of-the-art shows the effectiveness of the proposed approach.",
"Remote sensing image retrieval has many important applications in civilian and military fields, such as disaster monitoring and target detecting. However, the existing research on image retrieval, mainly including to two directions, text based and content based, cannot meet the rapid and convenient needs of some special applications and emergency scenes. Based on text, the retrieval is limited by keyboard inputting because of its lower efficiency for some urgent situations and based on content, it needs an example image as reference, which usually does not exist. Yet speech, as a direct, natural and efficient human-machine interactive way, can make up these shortcomings. Hence, a novel cross-modal retrieval method for remote sensing image and spoken audio is proposed in this paper. We first build a large-scale remote sensing image dataset with plenty of manual annotated spoken audio captions for the cross-modal retrieval task. Then a Deep Visual-Audio Network is designed to directly learn the correspondence of image and audio. And this model integrates feature extracting and multi-modal learning into the same network. Experiments on the proposed dataset verify the effectiveness of our approach and prove that it is feasible for speech-to-image retrieval.",
"",
"Predicting depth is an essential component in understanding the 3D geometry of a scene. While for stereo images local correspondence suffices for estimation, finding depth relations from a single image is less straightforward, requiring integration of both global and local information from various cues. Moreover, the task is inherently ambiguous, with a large source of uncertainty coming from the overall scale. In this paper, we present a new method that addresses this task by employing two deep network stacks: one that makes a coarse global prediction based on the entire image, and another that refines this prediction locally. We also apply a scale-invariant error to help measure depth relations rather than scale. By leveraging the raw datasets as large sources of training data, our method achieves state-of-the-art results on both NYU Depth and KITTI, and matches detailed depth boundaries without the need for superpixelation."
]
} |
1904.04794 | 2938495304 | We address the problem of cross-modal information retrieval in the domain of remote sensing. In particular, we are interested in two application scenarios: i) cross-modal retrieval between panchromatic (PAN) and multi-spectral imagery, and ii) multi-label image retrieval between very high resolution (VHR) images and speech based label annotations. Notice that these multi-modal retrieval scenarios are more challenging than the traditional uni-modal retrieval approaches given the inherent differences in distributions between the modalities. However, with the growing availability of multi-source remote sensing data and the scarcity of enough semantic annotations, the task of multi-modal retrieval has recently become extremely important. In this regard, we propose a novel deep neural network based architecture which is considered to learn a discriminative shared feature space for all the input modalities, suitable for semantically coherent information retrieval. Extensive experiments are carried out on the benchmark large-scale PAN - multi-spectral DSRSID dataset and the multi-label UC-Merced dataset. Together with the Merced dataset, we generate a corpus of speech signals corresponding to the labels. Superior performance with respect to the current state-of-the-art is observed in all the cases. | In contrast to the very few previous methods like @cite_11 @cite_15 in cross-modal retrieval in RS, we focus on the discriminative and domain independence of the shared latent feature space simultaneously. We find that our real-valued feature learning strategy outperforms the previous discriminative hashing based feature encoding substantially. Besides, we introduce the new notion of cross-modal retrieval between multi-label image and speech domains to aid the visually impaired. We also feel that the present works in cross-modal retrieval which have been done on image - text domain having a single label is effectively a classification problem as we have just a single label corresponding to any image. Hence we moved to a more challenging problem of image - speech. This can have a unique application to help a visually impaired person. For any image, a visually impaired person can have their corresponding audio label. Extensive experiments are performed to showcase the efficacy of the proposed framework. | {
"cite_N": [
"@cite_15",
"@cite_11"
],
"mid": [
"2898008292",
"2808376087"
],
"abstract": [
"Remote sensing image retrieval has many important applications in civilian and military fields, such as disaster monitoring and target detecting. However, the existing research on image retrieval, mainly including to two directions, text based and content based, cannot meet the rapid and convenient needs of some special applications and emergency scenes. Based on text, the retrieval is limited by keyboard inputting because of its lower efficiency for some urgent situations and based on content, it needs an example image as reference, which usually does not exist. Yet speech, as a direct, natural and efficient human-machine interactive way, can make up these shortcomings. Hence, a novel cross-modal retrieval method for remote sensing image and spoken audio is proposed in this paper. We first build a large-scale remote sensing image dataset with plenty of manual annotated spoken audio captions for the cross-modal retrieval task. Then a Deep Visual-Audio Network is designed to directly learn the correspondence of image and audio. And this model integrates feature extracting and multi-modal learning into the same network. Experiments on the proposed dataset verify the effectiveness of our approach and prove that it is feasible for speech-to-image retrieval.",
"Due to the urgent demand for remote sensing big data analysis, large-scale remote sensing image retrieval (LSRSIR) attracts increasing attention from researchers. Generally, LSRSIR can be divided into two categories as follows: uni-source LSRSIR (US-LSRSIR) and cross-source LSRSIR (CS-LSRSIR). More specifically, US-LSRSIR means the inquiry remote sensing image and images in the searching data set come from the same remote sensing data source, whereas CS-LSRSIR is designed to retrieve remote sensing images with a similar content to the inquiry remote sensing image that are from a different remote sensing data source. In the literature, US-LSRSIR has been widely exploited, but CS-LSRSIR is rarely discussed. In practical situations, remote sensing images from different kinds of remote sensing data sources are continually increasing, so there is a great motivation to exploit CS-LSRSIR. Therefore, this paper focuses on CS-LSRSIR. To cope with CS-LSRSIR, this paper proposes source-invariant deep hashing convolutional neural networks (SIDHCNNs), which can be optimized in an end-to-end manner using a series of well-designed optimization constraints. To quantitatively evaluate the proposed SIDHCNNs, we construct a dual-source remote sensing image data set that contains eight typical land-cover categories and 10 000 dual samples in each category. Extensive experiments show that the proposed SIDHCNNs can yield substantial improvements over several baselines involving the most recent techniques."
]
} |
1904.04466 | 2936096132 | Improving model performance is always the key problem in machine learning including deep learning. However, stand-alone neural networks always suffer from marginal effect when stacking more layers. At the same time, ensemble is a useful technique to further enhance model performance. Nevertheless, training several independent stand-alone deep neural networks costs multiple resources. In this work, we propose Intra-Ensemble, an end-to-end strategy with stochastic training operations to train several sub-networks simultaneously within one neural network. Additional parameter size is marginal since the majority of parameters are mutually shared. Meanwhile, stochastic training increases the diversity of sub-networks with weight sharing, which significantly enhances intra-ensemble performance. Extensive experiments prove the applicability of intra-ensemble on various kinds of datasets and network architectures. Our models achieve comparable results with the state-of-the-art architectures on CIFAR-10 and CIFAR-100. | Neural architecture search Recently, neural architecture search(NAS) @cite_30 @cite_25 @cite_42 has shown great improvement in designing smaller and more accurate network architectures. Under carefully designed search space and search algorithms, the searched architecture can attain a good trade-off between accuracy and latency. One-shot models @cite_0 @cite_36 train a cumbersome network with different optional operations each layer to approximately estimate the stand-alone model using its corresponding sub-network. This work inspires us to utilize the diversity and flexibility of one-shot model while maintaining the accuracy of its sub-networks. DARTS @cite_2 , ProxylessNas @cite_13 and FBNet @cite_47 propose differentiable strategies to implement efficient search of network architecture without any controller or hypernetwork. | {
"cite_N": [
"@cite_30",
"@cite_36",
"@cite_42",
"@cite_0",
"@cite_2",
"@cite_47",
"@cite_13",
"@cite_25"
],
"mid": [
"2885311373",
"",
"2553303224",
"2885820039",
"2951104886",
"2904699287",
"2902251695",
""
],
"abstract": [
"Deep Learning has enabled remarkable progress over the last years on a variety of tasks, such as image recognition, speech recognition, and machine translation. One crucial aspect for this progress are novel neural architectures. Currently employed architectures have mostly been developed manually by human experts, which is a time-consuming and error-prone process. Because of this, there is growing interest in automated neural architecture search methods. We provide an overview of existing work in this field of research and categorize them according to three dimensions: search space, search strategy, and performance estimation strategy.",
"",
"Neural networks are powerful and flexible models that work well for many difficult learning tasks in image, speech and natural language understanding. Despite their success, neural networks are still hard to design. In this paper, we use a recurrent network to generate the model descriptions of neural networks and train this RNN with reinforcement learning to maximize the expected accuracy of the generated architectures on a validation set. On the CIFAR-10 dataset, our method, starting from scratch, can design a novel network architecture that rivals the best human-invented architecture in terms of test set accuracy. Our CIFAR-10 model achieves a test error rate of 3.65, which is 0.09 percent better and 1.05x faster than the previous state-of-the-art model that used a similar architectural scheme. On the Penn Treebank dataset, our model can compose a novel recurrent cell that outperforms the widely-used LSTM cell, and other state-of-the-art baselines. Our cell achieves a test set perplexity of 62.4 on the Penn Treebank, which is 3.6 perplexity better than the previous state-of-the-art model. The cell can also be transferred to the character language modeling task on PTB and achieves a state-of-the-art perplexity of 1.214.",
"",
"This paper addresses the scalability challenge of architecture search by formulating the task in a differentiable manner. Unlike conventional approaches of applying evolution or reinforcement learning over a discrete and non-differentiable search space, our method is based on the continuous relaxation of the architecture representation, allowing efficient search of the architecture using gradient descent. Extensive experiments on CIFAR-10, ImageNet, Penn Treebank and WikiText-2 show that our algorithm excels in discovering high-performance convolutional architectures for image classification and recurrent architectures for language modeling, while being orders of magnitude faster than state-of-the-art non-differentiable techniques. Our implementation has been made publicly available to facilitate further research on efficient architecture search algorithms.",
"Designing accurate and efficient ConvNets for mobile devices is challenging because the design space is combinatorially large. Due to this, previous neural architecture search (NAS) methods are computationally expensive. ConvNet architecture optimality depends on factors such as input resolution and target devices. However, existing approaches are too expensive for case-by-case redesigns. Also, previous work focuses primarily on reducing FLOPs, but FLOP count does not always reflect actual latency. To address these, we propose a differentiable neural architecture search (DNAS) framework that uses gradient-based methods to optimize ConvNet architectures, avoiding enumerating and training individual architectures separately as in previous methods. FBNets, a family of models discovered by DNAS surpass state-of-the-art models both designed manually and generated automatically. FBNet-B achieves 74.1 top-1 accuracy on ImageNet with 295M FLOPs and 23.1 ms latency on a Samsung S8 phone, 2.4x smaller and 1.5x faster than MobileNetV2-1.3 with similar accuracy. Despite higher accuracy and lower latency than MnasNet, we estimate FBNet-B's search cost is 420x smaller than MnasNet's, at only 216 GPU-hours. Searched for different resolutions and channel sizes, FBNets achieve 1.5 to 6.4 higher accuracy than MobileNetV2. The smallest FBNet achieves 50.2 accuracy and 2.9 ms latency (345 frames per second) on a Samsung S8. Over a Samsung-optimized FBNet, the iPhone-X-optimized model achieves a 1.4x speedup on an iPhone X.",
"Neural architecture search (NAS) has a great impact by automatically designing effective neural network architectures. However, the prohibitive computational demand of conventional NAS algorithms (e.g. @math GPU hours) makes it difficult to search the architectures on large-scale tasks (e.g. ImageNet). Differentiable NAS can reduce the cost of GPU hours via a continuous representation of network architecture but suffers from the high GPU memory consumption issue (grow linearly w.r.t. candidate set size). As a result, they need to utilize tasks, such as training on a smaller dataset, or learning with only a few blocks, or training just for a few epochs. These architectures optimized on proxy tasks are not guaranteed to be optimal on the target task. In this paper, we present that can learn the architectures for large-scale target tasks and target hardware platforms. We address the high memory consumption issue of differentiable NAS and reduce the computational cost (GPU hours and GPU memory) to the same level of regular training while still allowing a large candidate set. Experiments on CIFAR-10 and ImageNet demonstrate the effectiveness of directness and specialization. On CIFAR-10, our model achieves 2.08 test error with only 5.7M parameters, better than the previous state-of-the-art architecture AmoebaNet-B, while using 6 @math fewer parameters. On ImageNet, our model achieves 3.1 better top-1 accuracy than MobileNetV2, while being 1.2 @math faster with measured GPU latency. We also apply ProxylessNAS to specialize neural architectures for hardware with direct hardware metrics (e.g. latency) and provide insights for efficient CNN architecture design.",
""
]
} |
1904.04466 | 2936096132 | Improving model performance is always the key problem in machine learning including deep learning. However, stand-alone neural networks always suffer from marginal effect when stacking more layers. At the same time, ensemble is a useful technique to further enhance model performance. Nevertheless, training several independent stand-alone deep neural networks costs multiple resources. In this work, we propose Intra-Ensemble, an end-to-end strategy with stochastic training operations to train several sub-networks simultaneously within one neural network. Additional parameter size is marginal since the majority of parameters are mutually shared. Meanwhile, stochastic training increases the diversity of sub-networks with weight sharing, which significantly enhances intra-ensemble performance. Extensive experiments prove the applicability of intra-ensemble on various kinds of datasets and network architectures. Our models achieve comparable results with the state-of-the-art architectures on CIFAR-10 and CIFAR-100. | Parameter sharing Guided by the thought of minimizing training cost, parameter sharing succeed in improving model performance while reducing model size. This idea has been used in some works on NAS and network morphism. ENAS @cite_46 and EPNAS @cite_3 increase training efficiency by sharing weights and One-shot @cite_0 uses shared weights cumbersome network with optional operations to estimate the performance of stand-alone child networks. Meanwhile, some works apply network morphism to transfer information between networks @cite_61 @cite_19 . Based on network morphism, @cite_48 search for well-performing CNN architectures based on a simple hill climbing procedure. Another inspiring work called slimmable neural networks @cite_37 shows that it is feasible to train several sub-networks within a single network without great loss of accuracy. In our work, we make use of its switchable techniques to train weight sharing sub-networks. | {
"cite_N": [
"@cite_61",
"@cite_37",
"@cite_48",
"@cite_3",
"@cite_0",
"@cite_19",
"@cite_46"
],
"mid": [
"2178031510",
"2905741102",
"2769653148",
"2949925172",
"2885820039",
"",
"2785366763"
],
"abstract": [
"We introduce techniques for rapidly transferring the information stored in one neural net into another neural net. The main purpose is to accelerate the training of a significantly larger neural net. During real-world workflows, one often trains very many different neural networks during the experimentation and design process. This is a wasteful process in which each new model is trained from scratch. Our Net2Net technique accelerates the experimentation process by instantaneously transferring the knowledge from a previous network to each new deeper or wider network. Our techniques are based on the concept of function-preserving transformations between neural network specifications. This differs from previous approaches to pre-training that altered the function represented by a neural net when adding layers to it. Using our knowledge transfer mechanism to add depth to Inception modules, we demonstrate a new state of the art accuracy rating on the ImageNet dataset.",
"We present a simple and general method to train a single neural network executable at different widths (number of channels in a layer), permitting instant and adaptive accuracy-efficiency trade-offs at runtime. Instead of training individual networks with different width configurations, we train a shared network with switchable batch normalization. At runtime, the network can adjust its width on the fly according to on-device benchmarks and resource constraints, rather than downloading and offloading different models. Our trained networks, named slimmable neural networks, achieve similar (and in many cases better) ImageNet classification accuracy than individually trained models of MobileNet v1, MobileNet v2, ShuffleNet and ResNet-50 at different widths respectively. We also demonstrate better performance of slimmable models compared with individual ones across a wide range of applications including COCO bounding-box object detection, instance segmentation and person keypoint detection without tuning hyper-parameters. Lastly we visualize and discuss the learned features of slimmable networks. Code and models are available at: this https URL",
"Neural networks have recently had a lot of success for many tasks. However, neural network architectures that perform well are still typically designed manually by experts in a cumbersome trial-and-error process. We propose a new method to automatically search for well-performing CNN architectures based on a simple hill climbing procedure whose operators apply network morphisms, followed by short optimization runs by cosine annealing. Surprisingly, this simple method yields competitive results, despite only requiring resources in the same order of magnitude as training a single network. E.g., on CIFAR-10, our method designs and trains networks with an error rate below 6 in only 12 hours on a single GPU; training for one day reduces this error further, to almost 5 .",
"This paper addresses the difficult problem of finding an optimal neural architecture design for a given image classification task. We propose a method that aggregates two main results of the previous state-of-the-art in neural architecture search. These are, appealing to the strong sampling efficiency of a search scheme based on sequential model-based optimization (SMBO), and increasing training efficiency by sharing weights among sampled architectures. Sequential search has previously demonstrated its capabilities to find state-of-the-art neural architectures for image classification. However, its computational cost remains high, even unreachable under modest computational settings. Affording SMBO with weight-sharing alleviates this problem. On the other hand, progressive search with SMBO is inherently greedy, as it leverages a learned surrogate function to predict the validation error of neural architectures. This prediction is directly used to rank the sampled neural architectures. We propose to attenuate the greediness of the original SMBO method by relaxing the role of the surrogate function so it predicts architecture sampling probability instead. We demonstrate with experiments on the CIFAR-10 dataset that our method, denominated Efficient progressive neural architecture search (EPNAS), leads to increased search efficiency, while retaining competitiveness of found architectures.",
"",
"",
"We propose Efficient Neural Architecture Search (ENAS), a fast and inexpensive approach for automatic model design. In ENAS, a controller learns to discover neural network architectures by searching for an optimal subgraph within a large computational graph. The controller is trained with policy gradient to select a subgraph that maximizes the expected reward on the validation set. Meanwhile the model corresponding to the selected subgraph is trained to minimize a canonical cross entropy loss. Thanks to parameter sharing between child models, ENAS is fast: it delivers strong empirical performances using much fewer GPU-hours than all existing automatic model design approaches, and notably, 1000x less expensive than standard Neural Architecture Search. On the Penn Treebank dataset, ENAS discovers a novel architecture that achieves a test perplexity of 55.8, establishing a new state-of-the-art among all methods without post-training processing. On the CIFAR-10 dataset, ENAS designs novel architectures that achieve a test error of 2.89 , which is on par with NASNet (, 2018), whose test error is 2.65 ."
]
} |
1904.04552 | 2938619957 | We approach video object segmentation (VOS) by splitting the task into two sub-tasks: bounding box level tracking, followed by bounding box segmentation. Following this paradigm, we present BoLTVOS (Box-Level Tracking for VOS), which consists of an R-CNN detector conditioned on the first-frame bounding box to detect the object of interest, a temporal consistency rescoring algorithm, and a Box2Seg network that converts bounding boxes to segmentation masks. BoLTVOS performs VOS using only the firstframe bounding box without the mask. We evaluate our approach on DAVIS 2017 and YouTube-VOS, and show that it outperforms all methods that do not perform first-frame fine-tuning. We further present BoLTVOS-ft, which learns to segment the object in question using the first-frame mask while it is being tracked, without increasing the runtime. BoLTVOS-ft outperforms PReMVOS, the previously best performing VOS method on DAVIS 2016 and YouTube-VOS, while running up to 45 times faster. Our bounding box tracker also outperforms all previous short-term and longterm trackers on the bounding box level tracking datasets OTB 2015 and LTB35. | Category 2 methods use the first-frame mask without fine-tuning @cite_44 @cite_16 @cite_49 @cite_36 @cite_38 @cite_56 @cite_4 . These methods run much faster, but do not achieve the same result quality. | {
"cite_N": [
"@cite_38",
"@cite_4",
"@cite_36",
"@cite_56",
"@cite_44",
"@cite_49",
"@cite_16"
],
"mid": [
"2799157347",
"2916797271",
"2889658408",
"2963253279",
"2964157492",
"2799239273",
"2962825871"
],
"abstract": [
"We present an efficient method for the semi-supervised video object segmentation. Our method achieves accuracy competitive with state-of-the-art methods while running in a fraction of time compared to others. To this end, we propose a deep Siamese encoder-decoder network that is designed to take advantage of mask propagation and object detection while avoiding the weaknesses of both approaches. Our network, learned through a two-stage training process that exploits both synthetic and real data, works robustly without any online learning or post-processing. We validate our method on four benchmark sets that cover single and multiple object segmentation. On all the benchmark sets, our method shows comparable accuracy while having the order of magnitude faster runtime. We also provide extensive ablation and add-on studies to analyze and evaluate our framework.",
"",
"Video object segmentation is challenging yet important in a wide variety of applications for video analysis. Recent works formulate video object segmentation as a prediction task using deep nets to achieve appealing state-of-the-art performance. Due to the formulation as a prediction task, most of these methods require fine-tuning during test time, such that the deep nets memorize the appearance of the objects of interest in the given video. However, fine-tuning is time-consuming and computationally expensive, hence the algorithms are far from real time. To address this issue, we develop a novel matching based algorithm for video object segmentation. In contrast to memorization based classification techniques, the proposed approach learns to match extracted features to a provided template without memorizing the appearance of the objects. We validate the effectiveness and the robustness of the proposed method on the challenging DAVIS-16, DAVIS-17, Youtube-Objects and JumpCut datasets. Extensive results show that our method achieves comparable performance without fine-tuning and is much more favorable in terms of computational time.",
"This paper tackles the task of semi-supervised video object segmentation, i.e., the separation of an object from the background in a video, given the mask of the first frame. We present One-Shot Video Object Segmentation (OSVOS), based on a fully-convolutional neural network architecture that is able to successively transfer generic semantic information, learned on ImageNet, to the task of foreground segmentation, and finally to learning the appearance of a single annotated object of the test sequence (hence one-shot). Although all frames are processed independently, the results are temporally coherent and stable. We perform experiments on two annotated video segmentation databases, which show that OSVOS is fast and improves the state of the art by a significant margin (79.8 vs 68.0 ).",
"This paper tackles the problem of video object segmentation, given some user annotation which indicates the object of interest. The problem is formulated as pixel-wise retrieval in a learned embedding space: we embed pixels of the same object instance into the vicinity of each other, using a fully convolutional network trained by a modified triplet loss as the embedding model. Then the annotated pixels are set as reference and the rest of the pixels are classified using a nearest-neighbor approach. The proposed method supports different kinds of user input such as segmentation mask in the first frame (semi-supervised scenario), or a sparse set of clicked points (interactive scenario). In the semi-supervised scenario, we achieve results competitive with the state of the art but at a fraction of computation cost (275 milliseconds per frame). In the interactive scenario where the user is able to refine their input iteratively, the proposed method provides instant response to each input, and reaches comparable quality to competing methods with much less interaction.",
"Online video object segmentation is a challenging task as it entails to process the image sequence timely and accurately. To segment a target object through the video, numerous CNN-based methods have been developed by heavily finetuning on the object mask in the first frame, which is time-consuming for online applications. In this paper, we propose a fast and accurate video object segmentation algorithm that can immediately start the segmentation process once receiving the images. We first utilize a part-based tracking method to deal with challenging factors such as large deformation, occlusion, and cluttered background. Based on the tracked bounding boxes of parts, we construct a region-of-interest segmentation network to generate part masks. Finally, a similarity-based scoring function is adopted to refine these object parts by comparing them to the visual information in the first frame. Our method performs favorably against state-of-the-art algorithms in accuracy on the DAVIS benchmark dataset, while achieving much faster runtime performance.",
"Video object segmentation targets segmenting a specific object throughout a video sequence when given only an annotated first frame. Recent deep learning based approaches find it effective to fine-tune a general-purpose segmentation model on the annotated frame using hundreds of iterations of gradient descent. Despite the high accuracy that these methods achieve, the fine-tuning process is inefficient and fails to meet the requirements of real world applications. We propose a novel approach that uses a single forward pass to adapt the segmentation model to the appearance of a specific object. Specifically, a second meta neural network named modulator is trained to manipulate the intermediate layers of the segmentation network given limited visual and spatial information of the target object. The experiments show that our approach is 70A— faster than fine-tuning approaches and achieves similar accuracy. Our model and code have been released at https: github.com linjieyangsc video_seg."
]
} |
1904.04552 | 2938619957 | We approach video object segmentation (VOS) by splitting the task into two sub-tasks: bounding box level tracking, followed by bounding box segmentation. Following this paradigm, we present BoLTVOS (Box-Level Tracking for VOS), which consists of an R-CNN detector conditioned on the first-frame bounding box to detect the object of interest, a temporal consistency rescoring algorithm, and a Box2Seg network that converts bounding boxes to segmentation masks. BoLTVOS performs VOS using only the firstframe bounding box without the mask. We evaluate our approach on DAVIS 2017 and YouTube-VOS, and show that it outperforms all methods that do not perform first-frame fine-tuning. We further present BoLTVOS-ft, which learns to segment the object in question using the first-frame mask while it is being tracked, without increasing the runtime. BoLTVOS-ft outperforms PReMVOS, the previously best performing VOS method on DAVIS 2016 and YouTube-VOS, while running up to 45 times faster. Our bounding box tracker also outperforms all previous short-term and longterm trackers on the bounding box level tracking datasets OTB 2015 and LTB35. | Visual Object Tracking (VOT). BoLTVOS splits the VOS task into box-level tracking and bounding box segmentation. Box-level tracking is commonly evaluated as the VOT task, which is similar to VOS, but only requires bounding box input and output instead of segmentation masks, and only deals with a single object. Recently, a number of long-term VOT benchmarks have been released which differ from traditional VOT in that the object to be tracked often disappears and reappears. The field of VOT has been driven by the success of a number of recent benchmarks such as the yearly VOT challenges @cite_51 @cite_48 , the Online Tracking Benchmark (OTB) @cite_26 @cite_55 , LTB35 @cite_48 and OxUvA @cite_41 for long-term VOT, and many others @cite_12 @cite_20 @cite_42 @cite_32 @cite_0 . | {
"cite_N": [
"@cite_26",
"@cite_41",
"@cite_48",
"@cite_55",
"@cite_42",
"@cite_32",
"@cite_0",
"@cite_20",
"@cite_51",
"@cite_12"
],
"mid": [
"2089961441",
"2794669410",
"2913466142",
"2158592639",
"2798799804",
"2518876086",
"2891033863",
"2898200825",
"2158827467",
"2794744029"
],
"abstract": [
"Object tracking is one of the most important components in numerous applications of computer vision. While much progress has been made in recent years with efforts on sharing code and datasets, it is of great importance to develop a library and benchmark to gauge the state of the art. After briefly reviewing recent advances of online object tracking, we carry out large scale experiments with various evaluation criteria to understand how these algorithms perform. The test image sequences are annotated with different attributes for performance evaluation and analysis. By analyzing quantitative results, we identify effective approaches for robust tracking and provide potential future research directions in this field.",
"We introduce the OxUvA dataset and benchmark for evaluating single-object tracking algorithms. Benchmarks have enabled great strides in the field of object tracking by defining standardized evaluations on large sets of diverse videos. However, these works have focused exclusively on sequences that are just tens of seconds in length and in which the target is always visible. Consequently, most researchers have designed methods tailored to this “short-term” scenario, which is poorly representative of practitioners’ needs. Aiming to address this disparity, we compile a long-term, large-scale tracking dataset of sequences with average length greater than two minutes and with frequent target object disappearance. The OxUvA dataset is much larger than the object tracking datasets of recent years: it comprises 366 sequences spanning 14 h of video. We assess the performance of several algorithms, considering both the ability to locate the target and to determine whether it is present or absent. Our goal is to offer the community a large and diverse benchmark to enable the design and evaluation of tracking methods ready to be used “in the wild”. The project website is oxuva.net.",
"The Visual Object Tracking challenge VOT2018 is the sixth annual tracker benchmarking activity organized by the VOT initiative. Results of over eighty trackers are presented; many are state-of-the-art trackers published at major computer vision conferences or in journals in the recent years. The evaluation included the standard VOT and other popular methodologies for short-term tracking analysis and a “real-time” experiment simulating a situation where a tracker processes images as if provided by a continuously running sensor. A long-term tracking subchallenge has been introduced to the set of standard VOT sub-challenges. The new subchallenge focuses on long-term tracking properties, namely coping with target disappearance and reappearance. A new dataset has been compiled and a performance evaluation methodology that focuses on long-term tracking capabilities has been adopted. The VOT toolkit has been updated to support both standard short-term and the new long-term tracking subchallenges. Performance of the tested trackers typically by far exceeds standard baselines. The source code for most of the trackers is publicly available from the VOT page. The dataset, the evaluation kit and the results are publicly available at the challenge website (http: votchallenge.net).",
"Object tracking has been one of the most important and active research areas in the field of computer vision. A large number of tracking algorithms have been proposed in recent years with demonstrated success. However, the set of sequences used for evaluation is often not sufficient or is sometimes biased for certain types of algorithms. Many datasets do not have common ground-truth object positions or extents, and this makes comparisons among the reported quantitative results difficult. In addition, the initial conditions or parameters of the evaluated tracking algorithms are not the same, and thus, the quantitative results reported in literature are incomparable or sometimes contradictory. To address these issues, we carry out an extensive evaluation of the state-of-the-art online object-tracking algorithms with various evaluation criteria to understand how these methods perform within the same framework. In this work, we first construct a large dataset with ground-truth object positions and extents for tracking and introduce the sequence attributes for the performance analysis. Second, we integrate most of the publicly available trackers into one code library with uniform input and output formats to facilitate large-scale performance evaluation. Third, we extensively evaluate the performance of 31 algorithms on 100 sequences with different initialization settings. By analyzing the quantitative results, we identify effective approaches for robust tracking and provide potential future research directions in this field.",
"In this paper we present a large-scale visual object detection and tracking benchmark, named VisDrone2018, aiming at advancing visual understanding tasks on the drone platform. The images and video sequences in the benchmark were captured over various urban suburban areas of 14 different cities across China from north to south. Specifically, VisDrone2018 consists of 263 video clips and 10,209 images (no overlap with video clips) with rich annotations, including object bounding boxes, object categories, occlusion, truncation ratios, etc. With intensive amount of effort, our benchmark has more than 2.5 million annotated instances in 179,264 images video frames. Being the largest such dataset ever published, the benchmark enables extensive evaluation and investigation of visual analysis algorithms on the drone platform. In particular, we design four popular tasks with the benchmark, including object detection in images, object detection in videos, single object tracking, and multi-object tracking. All these tasks are extremely challenging in the proposed dataset due to factors such as occlusion, large scale and pose variation, and fast motion. We hope the benchmark largely boost the research and development in visual analysis on drone platforms.",
"In this paper, we propose a new aerial video dataset and benchmark for low altitude UAV target tracking, as well as, a photo-realistic UAV simulator that can be coupled with tracking methods. Our benchmark provides the first evaluation of many state-of-the-art and popular trackers on 123 new and fully annotated HD video sequences captured from a low-altitude aerial perspective. Among the compared trackers, we determine which ones are the most suitable for UAV tracking both in terms of tracking accuracy and run-time. The simulator can be used to evaluate tracking algorithms in real-time scenarios before they are deployed on a UAV “in the field”, as well as, generate synthetic but photo-realistic tracking datasets with automatic ground truth annotations to easily extend existing real-world datasets. Both the benchmark and simulator are made publicly available to the vision community on our website to further research in the area of object tracking from UAVs. (https: ivul.kaust.edu.sa Pages pub-benchmark-simulator-uav.aspx.).",
"",
"In this work, we introduce a large high-diversity database for generic object tracking, called GOT-10k. GOT-10k is backboned by the semantic hierarchy of WordNet. It populates a majority of 563 object classes and 87 motion patterns in real-world, resulting in a scale of over 10 thousand video segments and 1.5 million bounding boxes. To our knowledge, GOT-10k is by far the richest motion trajectory dataset, and its coverage of object classes is more than a magnitude wider than similar scale counterparts. By publishing GOT-10k, we hope to encourage the development of generic purposed trackers that work for a wide range of moving objects and under diverse real-world scenarios. To promote generalization and avoid the evaluation results biased to seen classes, we follow the one-shot principle in dataset splitting where training and testing classes are zero-overlapped. We also carry out a series of analytical experiments to select a compact while highly representative testing subset -- it embodies 84 object classes and 32 motion patterns with only 180 video segments, allowing for efficient evaluation. Finally, we train and evaluate a number of representative trackers on GOT-10k and analyze their performance. The evaluation results suggest that tracking in real-world unconstrained videos is far from being solved, and only 40 of frames are successfully tracked using top ranking trackers. The database and toolkits are publicly available at this https URL.",
"This paper addresses the problem of single-target tracker performance evaluation. We consider the performance measures, the dataset and the evaluation system to be the most important components of tracker evaluation and propose requirements for each of them. The requirements are the basis of a new evaluation methodology that aims at a simple and easily interpretable tracker comparison. The ranking-based methodology addresses tracker equivalence in terms of statistical significance and practical differences. A fully-annotated dataset with per-frame annotations with several visual attributes is introduced. The diversity of its visual properties is maximized in a novel way by clustering a large number of videos according to their visual attributes. This makes it the most sophistically constructed and annotated dataset to date. A multi-platform evaluation system allowing easy integration of third-party trackers is presented as well. The proposed evaluation methodology was tested on the VOT2014 challenge on the new dataset and 38 trackers, making it the largest benchmark to date. Most of the tested trackers are indeed state-of-the-art since they outperform the standard baselines, resulting in a highly-challenging benchmark. An exhaustive analysis of the dataset from the perspective of tracking difficulty is carried out. To facilitate tracker comparison a new performance visualization technique is proposed.",
"Despite the numerous developments in object tracking, further improvement of current tracking algorithms is limited by small and mostly saturated datasets. As a matter of fact, data-hungry trackers based on deep-learning currently rely on object detection datasets due to the scarcity of dedicated large-scale tracking datasets. In this work, we present TrackingNet, the first large-scale dataset and benchmark for object tracking in the wild. We provide more than 30K videos with more than 14 million dense bounding box annotations. Our dataset covers a wide selection of object classes in broad and diverse context. By releasing such a large-scale dataset, we expect deep trackers to further improve and generalize. In addition, we introduce a new benchmark composed of 500 novel videos, modeled with a distribution similar to our training dataset. By sequestering the annotation of the test set and providing an online evaluation server, we provide a fair benchmark for future development of object trackers. Deep trackers fine-tuned on a fraction of our dataset improve their performance by up to 1.6 on OTB100 and up to 1.7 on TrackingNet Test. We provide an extensive benchmark on TrackingNet by evaluating more than 20 trackers. Our results suggest that object tracking in the wild is far from being solved."
]
} |
1904.04552 | 2938619957 | We approach video object segmentation (VOS) by splitting the task into two sub-tasks: bounding box level tracking, followed by bounding box segmentation. Following this paradigm, we present BoLTVOS (Box-Level Tracking for VOS), which consists of an R-CNN detector conditioned on the first-frame bounding box to detect the object of interest, a temporal consistency rescoring algorithm, and a Box2Seg network that converts bounding boxes to segmentation masks. BoLTVOS performs VOS using only the firstframe bounding box without the mask. We evaluate our approach on DAVIS 2017 and YouTube-VOS, and show that it outperforms all methods that do not perform first-frame fine-tuning. We further present BoLTVOS-ft, which learns to segment the object in question using the first-frame mask while it is being tracked, without increasing the runtime. BoLTVOS-ft outperforms PReMVOS, the previously best performing VOS method on DAVIS 2016 and YouTube-VOS, while running up to 45 times faster. Our bounding box tracker also outperforms all previous short-term and longterm trackers on the bounding box level tracking datasets OTB 2015 and LTB35. | The proposed BoLTVOS box-level tracker works by adapting an object detector for tracking by conditioning the detection on the given first-frame object template. Many previous methods also approach tracking as conditional detection. Most notably Siamese region proposal network methods such as SiamRPN @cite_29 which work by using a single stage RPN @cite_6 object detector and conditioning it on the first frame by cross-correlating the deep features of a local image patch with the deep features of the template. | {
"cite_N": [
"@cite_29",
"@cite_6"
],
"mid": [
"2799058067",
"2613718673"
],
"abstract": [
"Visual object tracking has been a fundamental topic in recent years and many deep learning based trackers have achieved state-of-the-art performance on multiple benchmarks. However, most of these trackers can hardly get top performance with real-time speed. In this paper, we propose the Siamese region proposal network (Siamese-RPN) which is end-to-end trained off-line with large-scale image pairs. Specifically, it consists of Siamese subnetwork for feature extraction and region proposal subnetwork including the classification branch and regression branch. In the inference phase, the proposed framework is formulated as a local one-shot detection task. We can pre-compute the template branch of the Siamese subnetwork and formulate the correlation layers as trivial convolution layers to perform online tracking. Benefit from the proposal refinement, traditional multi-scale test and online fine-tuning can be discarded. The Siamese-RPN runs at 160 FPS while achieving leading performance in VOT2015, VOT2016 and VOT2017 real-time challenges.",
"State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet [7] and Fast R-CNN [5] have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully-convolutional network that simultaneously predicts object bounds and objectness scores at each position. RPNs are trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. With a simple alternating optimization, RPN and Fast R-CNN can be trained to share convolutional features. For the very deep VGG-16 model [19], our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007 (73.2 mAP) and 2012 (70.4 mAP) using 300 proposals per image. Code is available at https: github.com ShaoqingRen faster_rcnn."
]
} |
1904.04363 | 2938060879 | Monitoring small objects against cluttered moving backgrounds is a huge challenge to future robotic vision systems. As a source of inspiration, insects are quite apt at searching for mates and tracking prey, which always appear as small dim speckles in the visual field. The exquisite sensitivity of insects for small target motion, as revealed recently, is coming from a class of specific neurons called small target motion detectors (STMDs). Although a few STMD-based models have been proposed, these existing models only use motion information for small target detection and cannot discriminate small targets from small-target-like background features (named fake features). To address this problem, this paper proposes a novel visual system model (STMD+) for small target motion detection, which is composed of four subsystems--ommatidia, motion pathway, contrast pathway, and mushroom body. Compared with the existing STMD-based models, the additional contrast pathway extracts directional contrast from luminance signals to eliminate false positive background motion. The directional contrast and the extracted motion information by the motion pathway are integrated into the mushroom body for small target discrimination. Extensive experiments showed the significant and consistent improvements of the proposed visual system model over the existing STMD-based models against fake features. | Small target motion detectors (STMDs) @cite_4 @cite_5 @cite_2 and lobula plate tangential cells (LPTCs) @cite_51 @cite_37 are widely investigated motion-sensitive neurons, where the former shows exquisite sensitivity to small target motion while the latter responds strongly to wide-field motion. Wiederman @cite_29 presented a mathematical model called ESTMD to simulate STMD neurons. It can detect the presence of small moving targets, but is unable to estimate motion direction. To address this issue, directional selectivity has been introduced into the ESTMD @cite_10 @cite_33 @cite_30 . However, these models cannot discriminate small targets from fake features, as they only make use of motion information. The first LPTC model called elementary motion detector (EMD) @cite_52 , is originally inferred from the insects' behavior. Following that, several studies have been done to further improve the EMD, such as @cite_3 @cite_23 @cite_19 . These models can detect all objects' motion, nevertheless they are unable to distinguish small moving objects from large ones. | {
"cite_N": [
"@cite_30",
"@cite_37",
"@cite_4",
"@cite_33",
"@cite_10",
"@cite_29",
"@cite_52",
"@cite_3",
"@cite_19",
"@cite_23",
"@cite_2",
"@cite_5",
"@cite_51"
],
"mid": [
"2963474154",
"2776678887",
"2586962938",
"2580553903",
"1981343219",
"1999508425",
"2062074945",
"2037283900",
"2029740022",
"2964275094",
"1991340096",
"2054077385",
"1545297250"
],
"abstract": [
"Discriminating targets moving against a cluttered background is a huge challenge, let alone detecting a target as small as one or a few pixels and tracking it in flight. In the insect's visual system, a class of specific neurons, called small target motion detectors (STMDs), have been identified as showing exquisite selectivity for small target motion. Some of the STMDs have also demonstrated direction selectivity which means these STMDs respond strongly only to their preferred motion direction. Direction selectivity is an important property of these STMD neurons which could contribute to tracking small targets such as mates in flight. However, little has been done on systematically modeling these directionally selective STMD neurons. In this paper, we propose a directionally selective STMD-based neural network for small target detection in a cluttered background. In the proposed neural network, a new correlation mechanism is introduced for direction selectivity via correlating signals relayed from two pixels. Then, a lateral inhibition mechanism is implemented on the spatial field for size selectivity of the STMD neurons. Finally, a population vector algorithm is used to encode motion direction of small targets. Extensive experiments showed that the proposed neural network not only is in accord with current biological findings, i.e., showing directional preferences but also worked reliably in detecting the small targets against cluttered backgrounds.",
"Neuronal representation and extraction of spatial information are essential for behavioral control. For flying insects, a plausible way to gain spatial information is to exploit distancedependent optic flow that is generated during translational self-motion. Optic flow is computed by arrays of local motion detectors retinotopically arranged in the second neuropile layer of the insect visual system. These motion detectors have adaptive response characteristics, i.e. their responses to motion with a constant or only slowly changing velocity decrease, while their sensitivity to rapid velocity changes is maintained or even increases. We analyzed by a modeling approach how motion adaptation affects signal representation at the output of arrays of motion detectors during simulated flight in artificial and natural 3D environments. We focused on translational flight, because spatial information is only contained in the optic flow induced by translational locomotion. Indeed, flies, bees and other insects segregate their flight into relatively long intersaccadic translational flight sections interspersed with brief and rapid saccadic turns, presumably to maximize periods of translation (80 of the flight). With a novel adaptive model of the insect visual motion pathway we could show that the motion detector responses to background structures of cluttered environments are largely attenuated as a consequence of motion adaptation, while responses to foreground objects stay constant or even increase. This conclusion even holds under the dynamic flight conditions of insects.",
"Summary Many animals rely on vision to detect objects such as conspecifics, predators, and prey. Hypercomplex cells found in feline cortex and small target motion detectors found in dragonfly and hoverfly optic lobes demonstrate robust tuning for small objects, with weak or no response to larger objects or movement of the visual panorama [1–3]. However, the relationship among anatomical, molecular, and functional properties of object detection circuitry is not understood. Here we characterize a specialized object detector in Drosophila , the lobula columnar neuron LC11 [4]. By imaging calcium dynamics with two-photon excitation microscopy, we show that LC11 responds to the omni-directional movement of a small object darker than the background, with little or no responses to static flicker, vertically elongated bars, or panoramic gratings. LC11 dendrites innervate multiple layers of the lobula, and each dendrite spans enough columns to sample 75° of visual space, yet the area that evokes calcium responses is only 20° wide and shows robust responses to a 2.2° object spanning less than half of one facet of the compound eye. The dendrites of neighboring LC11s encode object motion retinotopically, but the axon terminals fuse into a glomerular structure in the central brain where retinotopy is lost. Blocking inhibitory ionic currents abolishes small object sensitivity and facilitates responses to elongated bars and gratings. Our results reveal high-acuity object motion detection in the Drosophila optic lobe.",
"Robust and efficient target-tracking algorithms embedded on moving platforms, are a requirement for many computer vision and robotic applications. However, deployment of a real-time system is challenging, even with the computational power of modern hardware. As inspiration, we look to biological lightweight solutions-lightweight and low-powered flying insects. For example, dragonflies pursue prey and mates within cluttered, natural environments, deftly selecting their target amidst swarms. In our laboratory, we study the physiology and morphology of dragonfly 'small target motion detector' neurons likely to underlie this pursuit behaviour. Here we describe our insect-inspired tracking model derived from these data and compare its efficacy and efficiency with state-of-the-art engineering models. For model inputs, we use both publicly available video sequences, as well as our own task-specific dataset (small targets embedded within natural scenes). In the context of the tracking problem, we describe differences in object statistics within the video sequences. For the general dataset, our model often locks on to small components of larger objects, tracking these moving features. When input imagery includes small moving targets, for which our highly nonlinear filtering is matched, the robustness outperforms state-of-the-art trackers. In all scenarios, our insect-inspired tracker runs at least twice the speed of the comparison algorithms. (Less)",
"",
"We present a computational model for target discrimination based on intracellular recordings from neurons in the fly visual system. Determining how insects detect and track small moving features, often against cluttered moving backgrounds, is an intriguing challenge, both from a physiological and a computational perspective. Previous research has characterized higher-order neurons within the fly brain, known as ‘small target motion detectors’ (STMD), that respond robustly to moving features, even when the velocity of the target is matched to the background (i.e. with no relative motion cues). We recorded from intermediate-order neurons in the fly visual system that are well suited as a component along the target detection pathway. This full-wave rectifying, transient cell (RTC) reveals independent adaptation to luminance changes of opposite signs (suggesting separate ON and OFF channels) and fast adaptive temporal mechanisms, similar to other cell types previously described. From this physiological data we have created a numerical model for target discrimination. This model includes nonlinear filtering based on the fly optics, the photoreceptors, the 1st order interneurons (Large Monopolar Cells), and the newly derived parameters for the RTC. We show that our RTC-based target detection model is well matched to properties described for the STMDs, such as contrast sensitivity, height tuning and velocity tuning. The model output shows that the spatiotemporal profile of small targets is sufficiently rare within natural scene imagery to allow our highly nonlinear ‘matched filter’ to successfully detect most targets from the background. Importantly, this model can explain this type of feature discrimination without the need for relative motion cues.",
"",
"Summary Recent experiments have shown that motion detection in Drosophila starts with splitting the visual input into two parallel channels encoding brightness increments (ON) or decrements (OFF). This suggests the existence of either two (ON-ON, OFF-OFF) or four (for all pairwise interactions) separate motion detectors. To decide between these possibilities, we stimulated flies using sequences of ON and OFF brightness pulses while recording from motion-sensitive tangential cells. We found direction-selective responses to sequences of same sign (ON-ON, OFF-OFF), but not of opposite sign (ON-OFF, OFF-ON), refuting the existence of four separate detectors. Based on further measurements, we propose a model that reproduces a variety of additional experimental data sets, including ones that were previously interpreted as support for four separate detectors. Our experiments and the derived model mark an important step in guiding further dissection of the fly motion detection circuit.",
"Summary Many animals rely on visual motion detection for survival. Motion information is extracted from spatiotemporal intensity patterns on the retina, a paradigmatic neural computation. A phenomenological model, the Hassenstein-Reichardt correlator (HRC), relates visual inputs to neural activity and behavioral responses to motion, but the circuits that implement this computation remain unknown. By using cell-type specific genetic silencing, minimal motion stimuli, and in vivo calcium imaging, we examine two critical HRC inputs. These two pathways respond preferentially to light and dark moving edges. We demonstrate that these pathways perform overlapping but complementary subsets of the computations underlying the HRC. A numerical model implementing differential weighting of these operations displays the observed edge preferences. Intriguingly, these pathways are distinguished by their sensitivities to a stimulus correlation that corresponds to an illusory percept, \"reverse phi,\" that affects many species. Thus, this computational architecture may be widely used to achieve edge selectivity in motion detection.",
"A class of specialized neurons, called lobula plate tangential cells (LPTCs) has been shown to respond strongly to wide-field motion. The classic model, elementary motion detector (EMD) and its improved model, two-quadrant detector (TQD) have been proposed to simulate LPTCs. Although EMD and TQD can percept background motion, their outputs are so cluttered that it is difficult to discriminate actual motion direction of the background. In this paper, we propose a max operation mechanism to model a newly-found transmedullary neuron Tm9 whose physiological properties do not map onto EMD and TQD. This proposed max operation mechanism is able to improve the detection performance of TQD in cluttered background by filtering out irrelevant motion signals. We will demonstrate the functionality of this proposed mechanism in wide-field motion perception.",
"Despite being equipped with low-resolution eyes and tiny brains, many insects show exquisite abilities to detect and pursue targets even in highly textured surrounds. Target tracking behavior is subserved by neurons that are sharply tuned to the motion of small high-contrast targets. These neurons respond robustly to target motion, even against self-generated optic flow. A recent model, supported by neurophysiology, generates target selectivity by being sharply tuned to the unique spatiotemporal profile associated with target motion. Target neurons are likely connected in a complex network where some provide more direct output to behavior, whereas others serve an inter-regulatory role. These interactions may regulate attention and aid in the robust detection of targets in clutter observed in behavior.",
"Detection of targets that move within visual clutter is a common task for animals searching for prey or conspecifics, a task made even more difficult when a moving pursuer needs to analyze targets against the motion of background texture (clutter). Despite the limited optical acuity of the compound eye of insects, this challenging task seems to have been solved by their tiny visual system. Here we describe neurons found in the male hoverfly,Eristalis tenax, that respond selectively to small moving targets. Although many of these target neurons are inhibited by the motion of a background pattern, others respond to target motion within the receptive field under a surprisingly large range of background motion stimuli. Some neurons respond whether or not there is a speed differential between target and background. Analysis of responses to very small targets (smaller than the size of the visual field of single photoreceptors) or those targets with reduced contrast shows that these neurons have extraordinarily high contrast sensitivity. Our data suggest that rejection of background motion may result from extreme selectivity for small targets contrasting against local patches of the background, combined with this high sensitivity, such that background patterns rarely contain features that satisfactorily drive the neuron.",
"White noise techniques have been used widely to investigate sensory systems in both vertebrates and invertebrates. White noise stimuli are powerful in their ability to rapidly generate data that help the experimenter decipher the spatio-temporal dynamics of neural and behavioral responses. One type of white noise stimuli, maximal length shift register sequences (m-sequences), have recently become particularly popular for extracting response kernels in insect motion vision. We here use such m-sequences to extract the impulse responses to figure motion in hoverfly lobula plate tangential cells (LPTCs). Figure motion is behaviorally important and many visually guided animals orient towards salient features in the surround. We show that LPTCs respond robustly to figure motion in the receptive field. The impulse response is scaled down in amplitude when the figure size is reduced, but its time course remains unaltered. However, a low contrast stimulus generates a slower response with a significantly longer time-to-peak and half-width. Impulse responses in females have a slower time-to-peak than males, but are otherwise similar. Finally we show that the shapes of the impulse response to a figure and a widefield stimulus are very similar, suggesting that the figure response could be coded by the same input as the widefield response."
]
} |
1904.04363 | 2938060879 | Monitoring small objects against cluttered moving backgrounds is a huge challenge to future robotic vision systems. As a source of inspiration, insects are quite apt at searching for mates and tracking prey, which always appear as small dim speckles in the visual field. The exquisite sensitivity of insects for small target motion, as revealed recently, is coming from a class of specific neurons called small target motion detectors (STMDs). Although a few STMD-based models have been proposed, these existing models only use motion information for small target detection and cannot discriminate small targets from small-target-like background features (named fake features). To address this problem, this paper proposes a novel visual system model (STMD+) for small target motion detection, which is composed of four subsystems--ommatidia, motion pathway, contrast pathway, and mushroom body. Compared with the existing STMD-based models, the additional contrast pathway extracts directional contrast from luminance signals to eliminate false positive background motion. The directional contrast and the extracted motion information by the motion pathway are integrated into the mushroom body for small target discrimination. Extensive experiments showed the significant and consistent improvements of the proposed visual system model over the existing STMD-based models against fake features. | Traditional motion detection methods such as optical flow @cite_32 , background subtraction @cite_8 and temporal differencing @cite_26 , have been developed to detect normal-sized objects like pedestrians and vehicles. They utilize physical characteristics including shape, color and texture, to segment regions corresponding to moving objects from the background. Nonetheless, these methods would be powerless for objects that are as small as one pixel or a few pixels, because it is difficult to identify objects' physical characteristics in such small sizes. Additionally, the above-mentioned methods may not work for cluttered moving backgrounds, as small moving objects could be submerged among the pixel error when applying background motion compensation @cite_13 . | {
"cite_N": [
"@cite_26",
"@cite_13",
"@cite_32",
"@cite_8"
],
"mid": [
"2513283108",
"2072145684",
"2059612594",
"2618422320"
],
"abstract": [
"A rotation parameter extraction method based on temporal differencing and image edge detection from range-Doppler images is presented in this letter. The proposed method first detects the motion trail of the moving pixels caused by the rotating parts in temporal differential range-Doppler images using image edge detection. A Doppler-slow-time image is then generated from the edge pixels on the motion trail. Finally, the rotation parameters are extracted from the Doppler-slow-time image. The proposed method is simple, rapid, and practical. Computer simulations and experimental results demonstrate its effectiveness in terms of computation time compared with existing methods.",
"This paper proposes a new background subtraction method for detecting moving foreground objects from a nonstationary background. While background subtraction has traditionally worked well for a stationary background, the same cannot be implied for a nonstationary viewing sensor. To a limited extent, motion compensation for the nonstationary background can be applied. However, in practice, it is difficult to realize the motion compensation to sufficient pixel accuracy, and the traditional background subtraction algorithm will fail for a moving scene. The problem is further complicated when the moving target to be detected tracked is small, since the pixel error in motion that is compensating the background will subsume the small target. A spatial distribution of Gaussians (SDG) model is proposed to deal with moving object detection having motion compensation that is only approximately extracted. The distribution of each background pixel is temporally and spatially modeled. Based on this statistical model, a pixel in the current flame is then classified as belonging to the foreground or background. For this system to perform under lighting and environmental changes over an extended period of time, the background distribution must be updated with each incoming frame. A new background restoration and adaptation algorithm is developed for the nonstationary background. Test cases involving the detection of small moving objects within a highly textured background and with a pantilt tracking system are demonstrated successfully.",
"We propose a survey of optical flow estimation focusing on recent developments.We adopt a classification approach organizing methods in a comprehensive framework.The paper is conceived as a tutorial introducing and explaining the main concepts. Optical flow estimation is one of the oldest and still most active research domains in computer vision. In 35years, many methodological concepts have been introduced and have progressively improved performances, while opening the way to new challenges. In the last decade, the growing interest in evaluation benchmarks has stimulated a great amount of work. In this paper, we propose a survey of optical flow estimation classifying the main principles elaborated during this evolution, with a particular concern given to recent developments. It is conceived as a tutorial organizing in a comprehensive framework current approaches and practices. We give insights on the motivations, interests and limitations of modeling and optimization techniques, and we highlight similarities between methods to allow for a clear understanding of their behavior.",
"We propose an effective online background subtraction method, which can be robustly applied to practical videos that have variations in both foreground and background. Different from previous methods which often model the foreground as Gaussian or Laplacian distributions, we model the foreground for each frame with a specific mixture of Gaussians (MoG) distribution, which is updated online frame by frame. Particularly, our MoG model in each frame is regularized by the learned foreground background knowledge in previous frames. This makes our online MoG model highly robust, stable and adaptive to practical foreground and background variations. The proposed model can be formulated as a concise probabilistic MAP model, which can be readily solved by EM algorithm. We further embed an affine transformation operator into the proposed model, which can be automatically adjusted to fit a wide range of video background transformations and make the method more robust to camera movements. With using the sub-sampling technique, the proposed method can be accelerated to execute more than 250 frames per second on average, meeting the requirement of real-time background subtraction for practical video processing tasks. The superiority of the proposed method is substantiated by extensive experiments implemented on synthetic and real videos, as compared with state-of-the-art online and offline background subtraction methods."
]
} |
1904.04363 | 2938060879 | Monitoring small objects against cluttered moving backgrounds is a huge challenge to future robotic vision systems. As a source of inspiration, insects are quite apt at searching for mates and tracking prey, which always appear as small dim speckles in the visual field. The exquisite sensitivity of insects for small target motion, as revealed recently, is coming from a class of specific neurons called small target motion detectors (STMDs). Although a few STMD-based models have been proposed, these existing models only use motion information for small target detection and cannot discriminate small targets from small-target-like background features (named fake features). To address this problem, this paper proposes a novel visual system model (STMD+) for small target motion detection, which is composed of four subsystems--ommatidia, motion pathway, contrast pathway, and mushroom body. Compared with the existing STMD-based models, the additional contrast pathway extracts directional contrast from luminance signals to eliminate false positive background motion. The directional contrast and the extracted motion information by the motion pathway are integrated into the mushroom body for small target discrimination. Extensive experiments showed the significant and consistent improvements of the proposed visual system model over the existing STMD-based models against fake features. | Previous research and application of small target detection has mainly focused on infrared images @cite_34 @cite_41 @cite_27 . These infrared-based methods strongly rely on significant temperature differences between the background and objects of interest, such as rockets, jets and missiles. However, such significant temperature difference is rare in natural world. Moreover, the detection environment of these methods were mainly sky and or ocean, which are much more clear and homogeneous than the cluttered natural environments. These infrared-based methods may not work in a natural environment with lots of bushes, trees, sunlight and shadows, let alone to meet the needs of compact in size and low energy consumption in real applications @cite_40 @cite_9 @cite_46 @cite_47 @cite_31 . | {
"cite_N": [
"@cite_47",
"@cite_31",
"@cite_41",
"@cite_9",
"@cite_40",
"@cite_27",
"@cite_46",
"@cite_34"
],
"mid": [
"2591999327",
"2802636049",
"2341998679",
"2049989540",
"2523902637",
"2782405382",
"",
"1978993121"
],
"abstract": [
"This paper investigates adaptive fuzzy neural network (NN) control using impedance learning for a constrained robot, subject to unknown system dynamics, the effect of state constraints, and the uncertain compliant environment with which the robot comes into contact. A fuzzy NN learning algorithm is developed to identify the uncertain plant model. The prominent feature of the fuzzy NN is that there is no need to get the prior knowledge about the uncertainty and a sufficient amount of observed data. Also, impedance learning is introduced to tackle the interaction between the robot and its environment, so that the robot follows a desired destination generated by impedance learning. A barrier Lyapunov function is used to address the effect of state constraints. With the proposed control, the stability of the closed-loop system is achieved via Lyapunov’s stability theory, and the tracking performance is guaranteed under the condition of state constraints and uncertainty. Some simulation studies are carried out to illustrate the effectiveness of the proposed scheme.",
"Abstract Neural networks are applicable in many solutions for classification, prediction, control, etc. The variety of purposes is growing but with each new application the expectations are higher. We want neural networks to be more precise independently of the input data. Efficiency of the processing in a large manner depends on the training algorithm. Basically this procedure is based on the random selection of weights in which neurons connections are burdened. During training process we implement a method which involves modification of the weights to minimize the response error of the entire structure. Training continues until the minimum error value is reached — however in general the smaller it is, the time of weight modification is longer. Another problem is that training with the same set of data can cause different training times depending on the initial weight selection. To overcome arising problems we need a method that will boost the procedure and support final precision. In this article, we propose the use of multi-threading mechanism to minimize training time by rejecting unnecessary weights selection. In the mechanism we use a multi-core solution to select the best weights between all parallel trained networks. Proposed solution was tested for three types of neural networks (classic, sparking and convolutional) using sample classification problems. The results have shown positive aspects of the proposed idea: shorter training time and better efficiency in various tasks.",
"Infrared (IR) small target detection plays an important role in IR guidance systems. In this paper, a biologically inspired method called multiscale patch-based contrast measure (MPCM) is proposed for small IR target detection. MPCM can increase the contrast between target and background, which makes it easy to segment small target by simple adaptive thresholding method. Experimental results on three sequences demonstrate that the proposed small target detection method can not only suppress background clutter effectively even if with strong noise interference, but also detect targets accurately with low false alarm rate and high speed. HighlightsA biologically inspired target enhancement method called multiscale patch-based contrast measure (MPCM) is presented.Based on MPCM, a small IR target detection algorithm is designed.The proposed small target detection method achieves promising detection performance on three real IR image sequences.",
"Vision is one of the most useful sensory functions, but the real-time processing of the continuous, high-dimensional input signals provided by vision sensors is a major challenge in robot design. Conventional digital vision sensors tend to have excessive power consumption, size, and cost for useful applications. In this TechView, Indiveri and Douglas discuss an alternative strategy, namely neuromorphic vision sensors that are based on biological vision systems. In these sensors, specialized sensory processing functions inspired by biological systems such as fly eyes are integrated in parallel, asynchronous circuits that respond in real time. These sensors offer significant advantages over conventional vision sensors.",
"In this paper, we present a new bio-inspired vision system embedded for micro-robots. The vision system takes inspiration from locusts in detecting fast approaching objects. Neurophysiological research suggested that locusts use a wide-field visual neuron called lobula giant movement detector (LGMD) to respond to imminent collisions. In this paper, we present the implementation of the selected neuron model by a low-cost ARM processor as part of a composite vision module. As the first embedded LGMD vision module fits to a micro-robot, the developed system performs all image acquisition and processing independently. The vision module is placed on top of a micro-robot to initiate obstacle avoidance behavior autonomously. Both simulation and real-world experiments were carried out to test the reliability and robustness of the vision system. The results of the experiments with different scenarios demonstrated the potential of the bio-inspired vision system as a low-cost embedded module for autonomous robots.",
"Robust and effective detection of small target and false alarm (FA) suppression are the key techniques in infrared search and track systems. In this paper, the derivative entropy-based contrast measure (DECM) is proposed for small-target detection under various complex background clutters. Initially, different directional derivatives of an infrared image are calculated based on the facet model. Then, by analyzing the derivative properties of small target, the primitive entropy formula is improved by incorporating derivative information. With the improved entropy, the contrast measure is constructed to enhance small target and suppress background clutters in each derivative subband. Finally, the contrast measure maps derived from derivative subbands are fused together. The small target could be segmented easily from the fusion result. Experimental results demonstrate that DECM could effectively enhance dim small targets and suppress complex background clutters. Besides, DECM is also robust to infrared small-target images with noises of different levels. The detection results achieve higher detection ratio and lower FA compared with those of other methods under various infrared scenes.",
"",
"The robust detection of small targets is one of the key techniques in infrared search and tracking applications. A novel small target detection method in a single infrared image is proposed in this paper. Initially, the traditional infrared image model is generalized to a new infrared patch-image model using local patch construction. Then, because of the non-local self-correlation property of the infrared background image, based on the new model small target detection is formulated as an optimization problem of recovering low-rank and sparse matrices, which is effectively solved using stable principle component pursuit. Finally, a simple adaptive segmentation method is used to segment the target image and the segmentation result can be refined by post-processing. Extensive synthetic and real data experiments show that under different clutter backgrounds the proposed method not only works more stably for different target sizes and signal-to-clutter ratio values, but also has better detection performance compared with conventional baseline methods."
]
} |
1904.04306 | 2936238894 | A blockchain, during its lifetime, records large amounts of data, that in a common usage its kept on its entirety. In a robotics environment, the old information is useful for human evaluation, or oracles interfacing with the blockchain but it is not useful for the robots that require only current information in order to continue their work. This causes a storage problem in blockchain nodes that have limited storage capacity, such as in the case of nodes attached to robots that are usually built around embedded solutions. This paper presents a time-segmentation solution for devices with limited storage capacity, integrated in a particular robot-directed blockchain called RobotChain. Results are presented regarding the proposed solution that show that the goal of restricting each node's capacity is reached without compromising all the benefits that arise from the use of blockchains in these contexts, and on the contrary, it allows for cheap nodes to use this blockchain, reduces storage costs and allows faster deployment of new nodes. | Ferrer @cite_8 presents blockchain technology as a mean to improve robot swarms by solving problems, such as data confidentiality, distributed decision making, dynamic environment working capacity without master control program modification and a mean for legal responsibility for the members of the swarm. This last feature allows "integration" of robotic swarms with the human society. | {
"cite_N": [
"@cite_8"
],
"mid": [
"2484910796"
],
"abstract": [
"Swarms of robots will revolutionize many industrial applications, from targeted material delivery to precision farming. However, several of the heterogeneous characteristics that make them ideal for certain future applications—robot autonomy, decentralized control, collective emergent behavior, etc.—hinder the evolution of the technology from academic institutions to real-world problems. Blockchain, an emerging technology originated in the Bitcoin field, demonstrates that by combining peer-to-peer networks with cryptographic algorithms a group of agents can reach an agreement on a particular state of affairs and record that agreement without the need for a controlling authority. The combination of blockchain with other distributed systems, such as robotic swarm systems, can provide the necessary capabilities to make robotic swarm operations more secure, autonomous, flexible and even profitable. This position paper explains how blockchain technology can provide innovative solutions to four emergent issues in the swarm robotics research field. New security, decision making, behavior differentiation and business models for swarm robotic systems are described by providing case scenarios and examples. Finally, limitations and possible future problems that arise from the combination of these two technologies are described."
]
} |
1904.04306 | 2936238894 | A blockchain, during its lifetime, records large amounts of data, that in a common usage its kept on its entirety. In a robotics environment, the old information is useful for human evaluation, or oracles interfacing with the blockchain but it is not useful for the robots that require only current information in order to continue their work. This causes a storage problem in blockchain nodes that have limited storage capacity, such as in the case of nodes attached to robots that are usually built around embedded solutions. This paper presents a time-segmentation solution for devices with limited storage capacity, integrated in a particular robot-directed blockchain called RobotChain. Results are presented regarding the proposed solution that show that the goal of restricting each node's capacity is reached without compromising all the benefits that arise from the use of blockchains in these contexts, and on the contrary, it allows for cheap nodes to use this blockchain, reduces storage costs and allows faster deployment of new nodes. | In @cite_3 a proof-of-concept method is presented using blockchain smart contracts. These smart contracts allow improved security of the robotic swarm, improved the stability of the swarm coordination mechanisms and adds a way of expelling rogue members from the swarm. This concept has been studied for its performance in decision making regards of presence or absence of Byzantine robots. Byzantine Fault tolerance is the concern for fault-tolerance in distributed computer systems where components may fail, be unreliable, or be rogue agents. | {
"cite_N": [
"@cite_3"
],
"mid": [
"2808250907"
],
"abstract": [
"While swarm robotics systems are often claimed to be highly fault-tolerant, so far research has limited its attention to safe laboratory settings and has virtually ignored security issues in the presence of Byzantine robots---i.e., robots with arbitrarily faulty or malicious behavior. However, in many applications one or more Byzantine robots may suffice to let current swarm coordination mechanisms fail with unpredictable or disastrous outcomes. In this paper, we provide a proof-of-concept for managing security issues in swarm robotics systems via blockchain technology. Our approach uses decentralized programs executed via blockchain technology (blockchain-based smart contracts) to establish secure swarm coordination mechanisms and to identify and exclude Byzantine swarm members. We studied the performance of our blockchain-based approach in a collective decision-making scenario both in the presence and absence of Byzantine robots and compared our results to those obtained with an existing collective decision approach. The results show a clear advantage of the blockchain approach when Byzantine robots are part of the swarm."
]
} |
1904.04306 | 2936238894 | A blockchain, during its lifetime, records large amounts of data, that in a common usage its kept on its entirety. In a robotics environment, the old information is useful for human evaluation, or oracles interfacing with the blockchain but it is not useful for the robots that require only current information in order to continue their work. This causes a storage problem in blockchain nodes that have limited storage capacity, such as in the case of nodes attached to robots that are usually built around embedded solutions. This paper presents a time-segmentation solution for devices with limited storage capacity, integrated in a particular robot-directed blockchain called RobotChain. Results are presented regarding the proposed solution that show that the goal of restricting each node's capacity is reached without compromising all the benefits that arise from the use of blockchains in these contexts, and on the contrary, it allows for cheap nodes to use this blockchain, reduces storage costs and allows faster deployment of new nodes. | Danilov @cite_11 presents a trading model market named . It is focused on agent-based systems, where the behaviour is described as non-deterministic finite state automata, presenting a Model Checking verification technique in order to detect and filter malfunctioning agents. The validation method has the possibility to be implements on a consensus algorithm or as part of a blockchain application. Duckietown platform, with moving robots that follow a defined set of instructions, was used as a real life test. | {
"cite_N": [
"@cite_11"
],
"mid": [
"2801983037"
],
"abstract": [
"The decentralized trading market approach, where both autonomous agents and people can consume and produce services expanding own opportunities to reach goals, looks very promising as a part of the Fourth Industrial revolution. The key component of the approach is a blockchain platform that allows an interaction between agents via liability smart contracts. Reliability of a service provider is usually determined by a reputation model. However, this solution only warns future customers about an extent of trust to the service provider in case it could not execute any previous liabilities correctly. From the other hand a blockchain consensus protocol can additionally include a validation procedure that detects incorrect liability executions in order to suspend payment transactions to questionable service providers. The paper presents the validation methodology of a liability execution for agent-based service providers in a decentralized trading market, using the Model Checking method based on the mathematical model of finite state automata and Temporal Logic properties of interest. To demonstrate this concept, we implemented the methodology in the Duckietown application, moving an autonomous mobile robot to achieve a mission goal with the following behavior validation at the end of a completed scenario."
]
} |
1904.04306 | 2936238894 | A blockchain, during its lifetime, records large amounts of data, that in a common usage its kept on its entirety. In a robotics environment, the old information is useful for human evaluation, or oracles interfacing with the blockchain but it is not useful for the robots that require only current information in order to continue their work. This causes a storage problem in blockchain nodes that have limited storage capacity, such as in the case of nodes attached to robots that are usually built around embedded solutions. This paper presents a time-segmentation solution for devices with limited storage capacity, integrated in a particular robot-directed blockchain called RobotChain. Results are presented regarding the proposed solution that show that the goal of restricting each node's capacity is reached without compromising all the benefits that arise from the use of blockchains in these contexts, and on the contrary, it allows for cheap nodes to use this blockchain, reduces storage costs and allows faster deployment of new nodes. | In Ferrer @cite_4 , RoboChain is presented. It's a learning framework that attempts to solve various issues related to information sharing. These issues are privacy of personal information and machine learning data sharing, that allow multiple robots to work in different places while sharing their knowledge. This approach assumes a consortium blockchain where there is no public access and a level of trust between said parties is assumed. Since it is a consortium blockchain, the presence of a "malicious entity" is not considered, but it still provides a way to verify the integrity of the interactions and learning in the blockchain. | {
"cite_N": [
"@cite_4"
],
"mid": [
"2785967121"
],
"abstract": [
"This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No. 751615 and No. 701236."
]
} |
1904.04306 | 2936238894 | A blockchain, during its lifetime, records large amounts of data, that in a common usage its kept on its entirety. In a robotics environment, the old information is useful for human evaluation, or oracles interfacing with the blockchain but it is not useful for the robots that require only current information in order to continue their work. This causes a storage problem in blockchain nodes that have limited storage capacity, such as in the case of nodes attached to robots that are usually built around embedded solutions. This paper presents a time-segmentation solution for devices with limited storage capacity, integrated in a particular robot-directed blockchain called RobotChain. Results are presented regarding the proposed solution that show that the goal of restricting each node's capacity is reached without compromising all the benefits that arise from the use of blockchains in these contexts, and on the contrary, it allows for cheap nodes to use this blockchain, reduces storage costs and allows faster deployment of new nodes. | In @cite_9 , a blockchain framework for a secure ride-sharing service between autonomous vehicles and passengers is described. It uses a blockchain as a communication mechanism that is dependable and trustworthy. | {
"cite_N": [
"@cite_9"
],
"mid": [
"2808754907"
],
"abstract": [
"Car manufacturers are investing heavily in Autonomous Vehicles (AVs) because of their increasing popularity. AVs employ a combination of high-tech sensors and innovative algorithms to detect and respond to their surroundings, including radar, laser light, GPS, odometer, drive-by-wire control systems, and computer vision. When a user requests for a ride share or rent, the responding AV IoT device does not have the sufficient capability to store, process, or verify each other. Currently, every AV IoT device typically needs to rely on a trusted third party, which invokes another trust issue. If the trusted third party is rogue or loses credibility, the whole system will collapse. This paper introduces Chained of Things Framework and develops a blockchain-based secure ride-sharing service between passengers and autonomous vehicles. We implement an initial prototype and provide our initial results based on a proposed protocol."
]
} |
1904.04238 | 2939596636 | In this paper, we develop a novel Aligned-Spatial Graph Convolutional Network (ASGCN) model to learn effective features for graph classification. Our idea is to transform arbitrary-sized graphs into fixed-sized aligned grid structures, and define a new spatial graph convolution operation associated with the grid structures. We show that the proposed ASGCN model not only reduces the problems of information loss and imprecise information representation arising in existing spatially-based Graph Convolutional Network (GCN) models, but also bridges the theoretical gap between traditional Convolutional Neural Network (CNN) models and spatially-based GCN models. Moreover, the proposed ASGCN model can adaptively discriminate the importance between specified vertices during the process of spatial graph convolution, explaining the effectiveness of the proposed model. Experiments on standard graph datasets demonstrate the effectiveness of the proposed model. | Eq.) indicates that the spatial graph convolution operation of the DGCNN model cannot discriminate the importance between specified vertices in the convolution operation process. This is because the required filter weights @math are shared by each vertex, i.e., the feature transformations of the vertices are all based on the same trainable function. Thus, the DGCNN model cannot directly influence the aggregation process of the vertex features. In fact, this problem also arises in other spatially-based GCN models, e.g., the Neural Graph Fingerprint Network (NGFN) model @cite_13 , the Diffusion Convolution Neural Network (DCNN) model @cite_18 , etc. Since the associated spatial graph convolution operations of these models also take the similar form with that of the DGCNN model, i.e., the trainable parameters of their spatial graph convolution operations are also shared by each vertex. This drawback influences the effectiveness of the existing spatially-based GCN models for graph classification. In this paper, we aim to propose a new spatially-based GCN model to overcome the above problems. @math | {
"cite_N": [
"@cite_18",
"@cite_13"
],
"mid": [
"2963984147",
"2964113829"
],
"abstract": [
"We present diffusion-convolutional neural networks (DCNNs), a new model for graph-structured data. Through the introduction of a diffusion-convolution operation, we show how diffusion-based representations can be learned from graph-structured data and used as an effective basis for node classification. DCNNs have several attractive qualities, including a latent representation for graphical data that is invariant under isomorphism, as well as polynomial-time prediction and learning that can be represented as tensor operations and efficiently implemented on a GPU. Through several experiments with real structured datasets, we demonstrate that DCNNs are able to outperform probabilistic relational models and kernel-on-graph methods at relational node classification tasks.",
"We introduce a convolutional neural network that operates directly on graphs. These networks allow end-to-end learning of prediction pipelines whose inputs are graphs of arbitrary size and shape. The architecture we present generalizes standard molecular feature extraction methods based on circular fingerprints. We show that these data-driven features are more interpretable, and have better predictive performance on a variety of tasks."
]
} |
1904.04189 | 2930595145 | The task of temporally detecting and segmenting actions in untrimmed videos has seen an increased attention recently. One problem in this context arises from the need to define and label action boundaries to create annotations for training which is very time and cost intensive. To address this issue, we propose an unsupervised approach for learning action classes from untrimmed video sequences. To this end, we use a continuous temporal embedding of framewise features to benefit from the sequential nature of activities. Based on the latent space created by the embedding, we identify clusters of temporal segments across all videos that correspond to semantic meaningful action classes. The approach is evaluated on three challenging datasets, namely the Breakfast dataset, YouTube Instructions, and the 50Salads dataset. While previous works assumed that the videos contain the same high level activity, we furthermore show that the proposed approach can also be applied to a more general setting where the content of the videos is unknown. | There are also efforts for unsupervised learning of action classes. One of the first works that was tackling the problem of human motion segmentation without training data was proposed by Guerra-Filho and Aloimonos @cite_2 . They propose a basic segmentation with subsequent clustering based on sensory-motor data. Based on those representations, they propose the application of a parallel synchronous grammar system to learn atomic action representations similar to words in language. Another work in this context is proposed by Fox al @cite_33 where a Bayesian nonparametric approach helps to jointly model multiple related time series without further supervision. They apply their work on motion capture data. | {
"cite_N": [
"@cite_33",
"@cite_2"
],
"mid": [
"2167866158",
"2062465628"
],
"abstract": [
"We propose a Bayesian nonparametric approach to the problem of jointly modeling multiple related time series. Our model discovers a latent set of dynamical behaviors shared among the sequences, and segments each time series into regions defined by a subset of these behaviors. Using a beta process prior, the size of the behavior set and the sharing pattern are both inferred from data. We develop Markov chain Monte Carlo (MCMC) methods based on the Indian buffet process representation of the predictive distribution of the beta process. Our MCMC inference algorithm efficiently adds and removes behaviors via novel split-merge moves as well as data-driven birth and death proposals, avoiding the need to consider a truncated model. We demonstrate promising results on unsupervised segmentation of human motion capture data. 1. Introduction. Classical time series analysis has generally focused on the study of a single (potentially multivariate) time series. Instead, we consider analyzing collections of related time series, motivated by the increasing abundance of such data in many domains. In this work we explore this problem by considering time series produced by motion capture sensors on the joints of people performing exercise routines. An individual recording provides a multivariate time series that can be segmented into types of exercises (e.g., jumping jacks, arm-circles, and twists). Each exercise type describes locally coherent and simple dynamics that persist over a segment of time. We have such motion capture recordings from multiple individuals, each of whom performs some subset of a global set of exercises, as shown in Figure1. Our goal is to discover the set of global exercise types (“behaviors”) and their occurrences in each individual’s data stream. We would like to take advantage of the overlap between individuals: if a jumping-jack behavior is discovered in one sequence, then it can be used to model data for other individuals.",
"Human-centered computing (HCC) involves conforming computer technology to humans while naturally achieving human-machine interaction. In a human-centered system, the interaction focuses on human requirements, capabilities, and limitations. These anthropocentric systems also focus on the consideration of human sensory-motor skills in a wide range of activities. This ensures that the interface between artificial agents and human users accounts for perception and action in a novel interaction paradigm. In turn, this leads to behavior understanding through cognitive models that allow content description and, ultimately, the integration of real and virtual worlds. Our work focuses on building a language that maps to the lower-level sensory and motor languages and to the higher-level natural language. An empirically demonstrated human activity language provides sensory-motor-grounded representations for understanding human actions. A linguistic framework allows the analysis and synthesis of these actions."
]
} |
1904.04189 | 2930595145 | The task of temporally detecting and segmenting actions in untrimmed videos has seen an increased attention recently. One problem in this context arises from the need to define and label action boundaries to create annotations for training which is very time and cost intensive. To address this issue, we propose an unsupervised approach for learning action classes from untrimmed video sequences. To this end, we use a continuous temporal embedding of framewise features to benefit from the sequential nature of activities. Based on the latent space created by the embedding, we identify clusters of temporal segments across all videos that correspond to semantic meaningful action classes. The approach is evaluated on three challenging datasets, namely the Breakfast dataset, YouTube Instructions, and the 50Salads dataset. While previous works assumed that the videos contain the same high level activity, we furthermore show that the proposed approach can also be applied to a more general setting where the content of the videos is unknown. | In the context of video data, the temporal structure of video data has been exploited to fine-tune networks on training data without labels @cite_16 @cite_10 . The temporal ordering of video frames has also been used to learn feature representations for action recognition @cite_7 @cite_5 @cite_12 @cite_11 . Lee al @cite_7 learn a video representation in an unsupervised manner by solving a sequence sorting problem. Ramanathan al @cite_17 build their temporal embedding by leveraging contextual information of each frame on different resolution levels. Fernando al @cite_26 presented a method to capture the temporal evolution of actions based on frame appearance by learning a frame ranking function per video. In this way, they obtain a compact latent space for each video separately. A similar approach to learn a structured representation of postures and their temporal development was proposed by Milbich al @cite_37 . While these approaches address different tasks, Sener al @cite_31 proposed an unsupervised approach for learning action classes. They introduced an iterative approach which alternates between discriminative learning of the appearance of sub-activities from visual features and generative modeling of the temporal structure of sub-activities using a Generalized Mallows Model. | {
"cite_N": [
"@cite_37",
"@cite_26",
"@cite_7",
"@cite_17",
"@cite_5",
"@cite_31",
"@cite_16",
"@cite_10",
"@cite_12",
"@cite_11"
],
"mid": [
"2963437997",
"1926645898",
"2743563068",
"318042436",
"",
"2950816587",
"219040644",
"2745436507",
"",
""
],
"abstract": [
"Understanding human activity and being able to explain it in detail surpasses mere action classification by far in both complexity and value. The challenge is thus to describe an activity on the basis of its most fundamental constituents, the individual postures and their distinctive transitions. Supervised learning of such a fine-grained representation based on elementary poses is very tedious and does not scale. Therefore, we propose a completely unsupervised deep learning procedure based solely on video sequences, which starts from scratch without requiring pre-trained networks, predefined body models, or keypoints. A combinatorial sequence matching algorithm proposes relations between frames from subsets of the training data, while a CNN is reconciling the transitivity conflicts of the different subsets to learn a single concerted pose embedding despite changes in appearance across sequences. Without any manual annotation, the model learns a structured representation of postures and their temporal development. The model not only enables retrieval of similar postures but also temporal super-resolution. Additionally, based on a recurrent formulation, next frames can be synthesized.",
"In this paper we present a method to capture video-wide temporal information for action recognition. We postulate that a function capable of ordering the frames of a video temporally (based on the appearance) captures well the evolution of the appearance within the video. We learn such ranking functions per video via a ranking machine and use the parameters of these as a new video representation. The proposed method is easy to interpret and implement, fast to compute and effective in recognizing a wide variety of actions. We perform a large number of evaluations on datasets for generic action recognition (Hollywood2 and HMDB51), fine-grained actions (MPII- cooking activities) and gestures (Chalearn). Results show that the proposed method brings an absolute improvement of 7–10 , while being compatible with and complementary to further improvements in appearance and local motion based methods.",
"We present an unsupervised representation learning approach using videos without semantic labels. We leverage the temporal coherence as a supervisory signal by formulating representation learning as a sequence sorting task. We take temporally shuffled frames (i.e., in non-chronological order) as inputs and train a convolutional neural network to sort the shuffled sequences. Similar to comparison-based sorting algorithms, we propose to extract features from all frame pairs and aggregate them to predict the correct order. As sorting shuffled image sequence requires an understanding of the statistical temporal structure of images, training with such a proxy task allows us to learn rich and generalizable visual representation. We validate the effectiveness of the learned representation using our method as pre-training on high-level recognition problems. The experimental results show that our method compares favorably against state-of-the-art methods on action recognition, image classification and object detection tasks.",
"In this paper, we propose to learn temporal embeddings of video frames for complex video analysis. Large quantities of unlabeled video data can be easily obtained from the Internet. These videos possess the implicit weak label that they are sequences of temporally and semantically coherent images. We leverage this information to learn temporal embeddings for video frames by associating frames with the temporal context that they appear in. To do this, we propose a scheme for incorporating temporal context based on past and future frames in videos, and compare this to other contextual representations. In addition, we show how data augmentation using multi-resolution samples and hard negatives helps to significantly improve the quality of the learned embeddings. We evaluate various design decisions for learning temporal embeddings, and show that our embeddings can improve performance for multiple video tasks such as retrieval, classification, and temporal order recovery in unconstrained Internet video.",
"",
"This paper presents a new method for unsupervised segmentation of complex activities from video into multiple steps, or sub-activities, without any textual input. We propose an iterative discriminative-generative approach which alternates between discriminatively learning the appearance of sub-activities from the videos' visual features to sub-activity labels and generatively modelling the temporal structure of sub-activities using a Generalized Mallows Model. In addition, we introduce a model for background to account for frames unrelated to the actual activities. Our approach is validated on the challenging Breakfast Actions and Inria Instructional Videos datasets and outperforms both unsupervised and weakly-supervised state of the art.",
"Is strong supervision necessary for learning a good visual representation? Do we really need millions of semantically-labeled images to train a Convolutional Neural Network (CNN)? In this paper, we present a simple yet surprisingly powerful approach for unsupervised learning of CNN. Specifically, we use hundreds of thousands of unlabeled videos from the web to learn visual representations. Our key idea is that visual tracking provides the supervision. That is, two patches connected by a track should have similar visual representation in deep feature space since they probably belong to same object or object part. We design a Siamese-triplet network with a ranking loss function to train this CNN representation. Without using a single image from ImageNet, just using 100K unlabeled videos and the VOC 2012 dataset, we train an ensemble of unsupervised networks that achieves 52 mAP (no bounding box regression). This performance comes tantalizingly close to its ImageNet-supervised counterpart, an ensemble which achieves a mAP of 54.4 . We also show that our unsupervised network can perform competitively in other tasks such as surface-normal estimation.",
"Behavior analysis provides a crucial non-invasive and easily accessible diagnostic tool for biomedical research. A detailed analysis of posture changes during skilled motor tasks can reveal distinct functional deficits and their restoration during recovery. Our specific scenario is based on a neuroscientific study of rodents recovering from a large sensorimotor cortex stroke and skilled forelimb grasping is being recorded. Given large amounts of unlabeled videos that are recorded during such long-term studies, we seek an approach that captures fine-grained details of posture and its change during rehabilitation without costly manual supervision. Therefore, we utilize self-supervision to automatically learn accurate posture and behavior representations for analyzing motor function. Learning our model depends on the following fundamental elements: (i) limb detection based on a fully convolutional network is initialized solely using motion information, (ii) a novel self-supervised training of LSTMs using only temporal permutation yields a detailed representation of behavior, and (iii) back-propagation of this sequence representation also improves the description of individual postures. We establish a novel test dataset with expert annotations for evaluation of fine-grained behavior analysis. Moreover, we demonstrate the generality of our approach by successfully applying it to self-supervised learning of human posture on two standard benchmark datasets.",
"",
""
]
} |
1904.04443 | 2939038493 | An assumption widely used in recent neural style transfer methods is that image styles can be described by global statics of deep features like Gram or covariance matrices. Alternative approaches have represented styles by decomposing them into local pixel or neural patches. Despite the recent progress, most existing methods treat the semantic patterns of style image uniformly, resulting unpleasing results on complex styles. In this paper, we introduce a more flexible and general universal style transfer technique: multimodal style transfer (MST). MST explicitly considers the matching of semantic patterns in content and style images. Specifically, the style image features are clustered into sub-style components, which are matched with local content features under a graph cut formulation. A reconstruction network is trained to transfer each sub-style and render the final stylized result. We also generalize MST to improve some existing methods. Extensive experiments demonstrate the superior effectiveness, robustness, and flexibility of MST. | . Originating from non-realistic rendering @cite_34 , image style transfer is closely related to texture synthesis @cite_18 @cite_0 @cite_22 . Gatys @cite_27 were the first to formulate style transfer as the matching of multi-level deep features extracted from a pre-trained deep neural network. Lots of improvements have been proposed based on the works of Gatys @cite_27 . Johnson @cite_16 trained feed-forward style-specific network and produced one stylization with one model. Sanakoyeu @cite_6 further proposed a style-aware content loss for high-resolution style transfer. Jing @cite_14 proposed a StrokePyramid module to enable controllable stroke with adaptive receptive fields. However, these methods are either time consuming or have to re-train new models for new styles. | {
"cite_N": [
"@cite_18",
"@cite_14",
"@cite_22",
"@cite_16",
"@cite_6",
"@cite_0",
"@cite_27",
"@cite_34"
],
"mid": [
"2116013899",
"2962818016",
"2520715860",
"2331128040",
"2884041121",
"2964193438",
"2475287302",
"2109253138"
],
"abstract": [
"A non-parametric method for texture synthesis is proposed. The texture synthesis process grows a new image outward from an initial seed, one pixel at a time. A Markov random field model is assumed, and the conditional distribution of a pixel given all its neighbors synthesized so far is estimated by querying the sample image and finding all similar neighborhoods. The degree of randomness is controlled by a single perceptually intuitive parameter. The method aims at preserving as much local structure as possible and produces good results for a wide variety of synthetic and real-world textures.",
"The Fast Style Transfer methods have been recently proposed to transfer a photograph to an artistic style in real-time. This task involves controlling the stroke size in the stylized results, which remains an open challenge. In this paper, we present a stroke controllable style transfer network that can achieve continuous and spatial stroke size control. By analyzing the factors that influence the stroke size, we propose to explicitly account for the receptive field and the style image scales. We propose a StrokePyramid module to endow the network with adaptive receptive fields, and two training strategies to achieve faster convergence and augment new stroke sizes upon a trained model respectively. By combining the proposed runtime control strategies, our network can achieve continuous changes in stroke sizes and produce distinct stroke sizes in different spatial regions within the same output image.",
"Style transfer is a process of migrating a style from a given image to the content of another, synthesizing a new image, which is an artistic mixture of the two. Recent work on this problem adopting convolutional neural-networks (CNN) ignited a renewed interest in this field, due to the very impressive results obtained. There exists an alternative path toward handling the style transfer task, via the generalization of texture synthesis algorithms. This approach has been proposed over the years, but its results are typically less impressive compared with the CNN ones. In this paper, we propose a novel style transfer algorithm that extends the texture synthesis work of (2005), while aiming to get stylized images that are closer in quality to the CNN ones. We modify Kwatra’s algorithm in several key ways in order to achieve the desired transfer, with emphasis on a consistent way for keeping the content intact in selected regions, while producing hallucinated and rich style in others. The results obtained are visually pleasing and diverse, shown to be competitive with the recent CNN style transfer algorithms. The proposed algorithm is fast and flexible, being able to process any pair of content + style images.",
"We consider image transformation problems, where an input image is transformed into an output image. Recent methods for such problems typically train feed-forward convolutional neural networks using a per-pixel loss between the output and ground-truth images. Parallel work has shown that high-quality images can be generated by defining and optimizing perceptual loss functions based on high-level features extracted from pretrained networks. We combine the benefits of both approaches, and propose the use of perceptual loss functions for training feed-forward networks for image transformation tasks. We show results on image style transfer, where a feed-forward network is trained to solve the optimization problem proposed by in real-time. Compared to the optimization-based method, our network gives similar qualitative results but is three orders of magnitude faster. We also experiment with single-image super-resolution, where replacing a per-pixel loss with a perceptual loss gives visually pleasing results.",
"Recently, style transfer has received a lot of attention. While much of this research has aimed at speeding up processing, the approaches are still lacking from a principled, art historical standpoint: a style is more than just a single image or an artist, but previous work is limited to only a single instance of a style or shows no benefit from more images. Moreover, previous work has relied on a direct comparison of art in the domain of RGB images or on CNNs pre-trained on ImageNet, which requires millions of labeled object bounding boxes and can introduce an extra bias, since it has been assembled without artistic consideration. To circumvent these issues, we propose a style-aware content loss, which is trained jointly with a deep encoder-decoder network for real-time, high-resolution stylization of images and videos. We propose a quantitative measure for evaluating the quality of a stylized image and also have art historians rank patches from our approach against those from previous work. These and our qualitative results ranging from small image patches to megapixel stylistic images and videos show that our approach better captures the subtle nature in which a style affects content.",
"Here we introduce a new model of natural textures based on the feature spaces of convolutional neural networks optimised for object recognition. Samples from the model are of high perceptual quality demonstrating the generative power of neural networks trained in a purely discriminative fashion. Within the model, textures are represented by the correlations between feature maps in several layers of the network. We show that across layers the texture representations increasingly capture the statistical properties of natural images while making object information more and more explicit. The model provides a new tool to generate stimuli for neuroscience and might offer insights into the deep representations learned by convolutional neural networks.",
"Rendering the semantic content of an image in different styles is a difficult image processing task. Arguably, a major limiting factor for previous approaches has been the lack of image representations that explicitly represent semantic information and, thus, allow to separate image content from style. Here we use image representations derived from Convolutional Neural Networks optimised for object recognition, which make high level image information explicit. We introduce A Neural Algorithm of Artistic Style that can separate and recombine the image content and style of natural images. The algorithm allows us to produce new images of high perceptual quality that combine the content of an arbitrary photograph with the appearance of numerous wellknown artworks. Our results provide new insights into the deep image representations learned by Convolutional Neural Networks and demonstrate their potential for high level image synthesis and manipulation.",
"This paper surveys the field of nonphotorealistic rendering (NPR), focusing on techniques for transforming 2D input (images and video) into artistically stylized renderings. We first present a taxonomy of the 2D NPR algorithms developed over the past two decades, structured according to the design characteristics and behavior of each technique. We then describe a chronology of development from the semiautomatic paint systems of the early nineties, through to the automated painterly rendering systems of the late nineties driven by image gradient analysis. Two complementary trends in the NPR literature are then addressed, with reference to our taxonomy. First, the fusion of higher level computer vision and NPR, illustrating the trends toward scene analysis to drive artistic abstraction and diversity of style. Second, the evolution of local processing approaches toward edge-aware filtering for real-time stylization of images and video. The survey then concludes with a discussion of open challenges for 2D NPR identified in recent NPR symposia, including topics such as user and aesthetic evaluation."
]
} |
1904.04443 | 2939038493 | An assumption widely used in recent neural style transfer methods is that image styles can be described by global statics of deep features like Gram or covariance matrices. Alternative approaches have represented styles by decomposing them into local pixel or neural patches. Despite the recent progress, most existing methods treat the semantic patterns of style image uniformly, resulting unpleasing results on complex styles. In this paper, we introduce a more flexible and general universal style transfer technique: multimodal style transfer (MST). MST explicitly considers the matching of semantic patterns in content and style images. Specifically, the style image features are clustered into sub-style components, which are matched with local content features under a graph cut formulation. A reconstruction network is trained to transfer each sub-style and render the final stylized result. We also generalize MST to improve some existing methods. Extensive experiments demonstrate the superior effectiveness, robustness, and flexibility of MST. | The first arbitrary style transfer was proposed by Chen and Schmidt @cite_23 , who matched each content patch to the most similar style patch and swapped them. Luan @cite_30 proposed deep photo style transfer by adding a regularization term to the optimization function. Based on markov random field (MRF), Li and Wand @cite_9 proposed CNNMRF to enforce local patterns in deep feature space. Ruder @cite_31 improved video stylization with temporal coherence. Although their visual stylizations for arbitrary style are appealing, the results are not stable @cite_31 . | {
"cite_N": [
"@cite_30",
"@cite_9",
"@cite_31",
"@cite_23"
],
"mid": [
"2604721644",
"2275363859",
"2344328033",
"2564755245"
],
"abstract": [
"This paper introduces a deep-learning approach to photographic style transfer that handles a large variety of image content while faithfully transferring the reference style. Our approach builds upon the recent work on painterly transfer that separates style from the content of an image by considering different layers of a neural network. However, as is, this approach is not suitable for photorealistic style transfer. Even when both the input and reference images are photographs, the output still exhibits distortions reminiscent of a painting. Our contribution is to constrain the transformation from the input to the output to be locally affine in colorspace, and to express this constraint as a custom fully differentiable energy term. We show that this approach successfully suppresses distortion and yields satisfying photorealistic style transfers in a broad variety of scenarios, including transfer of the time of day, weather, season, and artistic edits.",
"This paper studies a combination of generative Markov random field (MRF) models and discriminatively trained deep convolutional neural networks (dCNNs) for synthesizing 2D images. The generative MRF acts on higher-levels of a dCNN feature pyramid, controling the image layout at an abstract level. We apply the method to both photographic and non-photo-realistic (artwork) synthesis tasks. The MRF regularizer prevents over-excitation artifacts and reduces implausible feature mixtures common to previous dCNN inversion approaches, permitting synthezing photographic content with increased visual plausibility. Unlike standard MRF-based texture synthesis, the combined system can both match and adapt local features with considerable variability, yielding results far out of reach of classic generative MRF methods.",
"In the past, manually re-drawing an image in a certain artistic style required a professional artist and a long time. Doing this for a video sequence single-handed was beyond imagination. Nowadays computers provide new possibilities. We present an approach that transfers the style from one image (for example, a painting) to a whole video sequence. We make use of recent advances in style transfer in still images and propose new initializations and loss functions applicable to videos. This allows us to generate consistent and stable stylized video sequences, even in cases with large motion and strong occlusion. We show that the proposed method clearly outperforms simpler baselines both qualitatively and quantitatively.",
"Artistic style transfer is an image synthesis problem where the content of an image is reproduced with the style of another. Recent works show that a visually appealing style transfer can be achieved by using the hidden activations of a pretrained convolutional neural network. However, existing methods either apply (i) an optimization procedure that works for any style image but is very expensive, or (ii) an efficient feedforward network that only allows a limited number of trained styles. In this work we propose a simpler optimization objective based on local matching that combines the content structure and style textures in a single layer of the pretrained network. We show that our objective has desirable properties such as a simpler optimization landscape, intuitive parameter tuning, and consistent frame-by-frame performance on video. Furthermore, we use 80,000 natural images and 80,000 paintings to train an inverse network that approximates the result of the optimization. This results in a procedure for artistic style transfer that is efficient but also allows arbitrary content and style images."
]
} |
1904.04443 | 2939038493 | An assumption widely used in recent neural style transfer methods is that image styles can be described by global statics of deep features like Gram or covariance matrices. Alternative approaches have represented styles by decomposing them into local pixel or neural patches. Despite the recent progress, most existing methods treat the semantic patterns of style image uniformly, resulting unpleasing results on complex styles. In this paper, we introduce a more flexible and general universal style transfer technique: multimodal style transfer (MST). MST explicitly considers the matching of semantic patterns in content and style images. Specifically, the style image features are clustered into sub-style components, which are matched with local content features under a graph cut formulation. A reconstruction network is trained to transfer each sub-style and render the final stylized result. We also generalize MST to improve some existing methods. Extensive experiments demonstrate the superior effectiveness, robustness, and flexibility of MST. | Recently, Huang @cite_17 proposed real-time style transfer by matching the mean-variance statistics between content and style features. Li @cite_29 further introduced whitening and coloring (WCT) by matching the covariance matrices. Li boosted style transfer with linear style transfer (LST) @cite_21 . Gu @cite_33 proposed deep feature reshuffle (DFR), which connects both local and global style losses used in parametric and non-parametric methods. Sheng @cite_5 proposed AvatarNet to enable multi-scale transfer for arbitrary style. Shen @cite_2 built meta networks by taking style images as inputs and generating corresponding image transformation networks directly. Mechrez @cite_10 proposed contextual loss for image transformation. However, these methods fail to treat the style patterns distinctively and neglect to adaptively match style patterns with content semantic information. For more neural style transfer works, readers can refer to the survey @cite_32 . | {
"cite_N": [
"@cite_33",
"@cite_29",
"@cite_21",
"@cite_32",
"@cite_2",
"@cite_5",
"@cite_10",
"@cite_17"
],
"mid": [
"2798520250",
"2962772087",
"2886566843",
"2613099748",
"2798454267",
"",
"2964309429",
"2603777577"
],
"abstract": [
"This paper introduces a novel method by reshuffling deep features (i.e., permuting the spacial locations of a feature map) of the style image for arbitrary style transfer. We theoretically prove that our new style loss based on reshuffle connects both global and local style losses respectively used by most parametric and non-parametric neural style transfer methods. This simple idea can effectively address the challenging issues in existing style transfer methods. On one hand, it can avoid distortions in local style patterns, and allow semantic-level transfer, compared with neural parametric methods. On the other hand, it can preserve globally similar appearance to the style image, and avoid wash-out artifacts, compared with neural non-parametric methods. Based on the proposed loss, we also present a progressive feature-domain optimization approach. The experiments show that our method is widely applicable to various styles, and produces better quality than existing methods.",
"Universal style transfer aims to transfer arbitrary visual styles to content images. Existing feed-forward based methods, while enjoying the inference efficiency, are mainly limited by inability of generalizing to unseen styles or compromised visual quality. In this paper, we present a simple yet effective method that tackles these limitations without training on any pre-defined styles. The key ingredient of our method is a pair of feature transforms, whitening and coloring, that are embedded to an image reconstruction network. The whitening and coloring transforms reflect direct matching of feature covariance of the content image to a given style image, which shares similar spirits with the optimization of Gram matrix based cost in neural style transfer. We demonstrate the effectiveness of our algorithm by generating high-quality stylized images with comparisons to a number of recent methods. We also analyze our method by visualizing the whitened features and synthesizing textures by simple feature coloring.",
"Given a random pair of images, an arbitrary style transfer method extracts the feel from the reference image to synthesize an output based on the look of the other content image. Recent arbitrary style transfer methods transfer second order statistics from reference image onto content image via a multiplication between content image features and a transformation matrix, which is computed from features with a pre-determined algorithm. These algorithms either require computationally expensive operations, or fail to model the feature covariance and produce artifacts in synthesized images. Generalized from these methods, in this work, we derive the form of transformation matrix theoretically and present an arbitrary style transfer approach that learns the transformation matrix with a feed-forward network. Our algorithm is highly efficient yet allows a flexible combination of multi-level styles while preserving content affinity during style transfer process. We demonstrate the effectiveness of our approach on four tasks: artistic style transfer, video and photo-realistic style transfer as well as domain adaptation, including comparisons with the state-of-the-art methods.",
"The seminal work of demonstrated the power of Convolutional Neural Networks (CNNs) in creating artistic imagery by separating and recombining image content and style. This process of using CNNs to render a content image in different styles is referred to as Neural Style Transfer (NST). Since then, NST has become a trending topic both in academic literature and industrial applications. It is receiving increasing attention and a variety of approaches are proposed to either improve or extend the original NST algorithm. In this paper, we aim to provide a comprehensive overview of the current progress towards NST. We first propose a taxonomy of current algorithms in the field of NST. Then, we present several evaluation methods and compare different NST algorithms both qualitatively and quantitatively. The review concludes with a discussion of various applications of NST and open problems for future research. A list of papers discussed in this review, corresponding codes, pre-trained models and more comparison results are publicly available at this https URL.",
"In this paper we propose a noval method to generate the specified network parameters through one feed-forward propagation in the meta networks for neural style transfer. Recent works on style transfer typically need to train image transformation networks for every new style, and the style is encoded in the network parameters by enormous iterations of stochastic gradient descent, which lacks the generalization ability to new style in the inference stage. To tackle these issues, we build a meta network which takes in the style image and generates a corresponding image transformation network directly. Compared with optimization-based methods for every style, our meta networks can handle an arbitrary new style within 19 milliseconds on one modern GPU card. The fast image transformation network generated by our meta network is only 449 KB, which is capable of real-time running on a mobile device. We also investigate the manifold of the style transfer networks by operating the hidden features from meta networks. Experiments have well validated the effectiveness of our method. Code and trained models will be released.",
"",
"Feed-forward CNNs trained for image transformation problems rely on loss functions that measure the similarity between the generated image and a target image. Most of the common loss functions assume that these images are spatially aligned and compare pixels at corresponding locations. However, for many tasks, aligned training pairs of images will not be available. We present an alternative loss function that does not require alignment, thus providing an effective and simple solution for a new space of problems. Our loss is based on both context and semantics – it compares regions with similar semantic meaning, while considering the context of the entire image. Hence, for example, when transferring the style of one face to another, it will translate eyes-to-eyes and mouth-to-mouth. Our code can be found at https: www.github.com roimehrez contextualLoss.",
"recently introduced a neural algorithm that renders a content image in the style of another image, achieving so-called style transfer. However, their framework requires a slow iterative optimization process, which limits its practical application. Fast approximations with feed-forward neural networks have been proposed to speed up neural style transfer. Unfortunately, the speed improvement comes at a cost: the network is usually tied to a fixed set of styles and cannot adapt to arbitrary new styles. In this paper, we present a simple yet effective approach that for the first time enables arbitrary style transfer in real-time. At the heart of our method is a novel adaptive instance normalization (AdaIN) layer that aligns the mean and variance of the content features with those of the style features. Our method achieves speed comparable to the fastest existing approach, without the restriction to a pre-defined set of styles. In addition, our approach allows flexible user controls such as content-style trade-off, style interpolation, color & spatial controls, all using a single feed-forward neural network."
]
} |
1904.04443 | 2939038493 | An assumption widely used in recent neural style transfer methods is that image styles can be described by global statics of deep features like Gram or covariance matrices. Alternative approaches have represented styles by decomposing them into local pixel or neural patches. Despite the recent progress, most existing methods treat the semantic patterns of style image uniformly, resulting unpleasing results on complex styles. In this paper, we introduce a more flexible and general universal style transfer technique: multimodal style transfer (MST). MST explicitly considers the matching of semantic patterns in content and style images. Specifically, the style image features are clustered into sub-style components, which are matched with local content features under a graph cut formulation. A reconstruction network is trained to transfer each sub-style and render the final stylized result. We also generalize MST to improve some existing methods. Extensive experiments demonstrate the superior effectiveness, robustness, and flexibility of MST. | . Many problems that arose in early vision can be naturally expressed in terms of energy minimization. For example, a large number of computer vision problems attempt to assign labels to pixels based on noisy measurements. Graph cuts is a powerful method to solve such discrete optimization problems. Greig @cite_36 were firstly successful solving graph cuts by using powerful min-cut max-flow algorithms from combinatorial optimization. Roy and Cox @cite_12 were the first to use these techniques for multi-camera stereo computation. Later, a growing number of researches in computer vision use graph-based energy minimization for a wide range of applications, which includes stereo @cite_8 , texture synthesis @cite_24 , image segmentation @cite_25 , object recognition @cite_28 , and others. In this paper, we formulate the matching between content and style features as an energy minimization problem. We approximate its global minimum via efficient graph cuts algorithms. To the best of our knowledge, we are the first to formulate style matching as energy minimization problem and solve it via graph cuts. | {
"cite_N": [
"@cite_8",
"@cite_36",
"@cite_28",
"@cite_24",
"@cite_25",
"@cite_12"
],
"mid": [
"1781337833",
"79315950",
"2126911940",
"2077786999",
"2126140058",
"1649464328"
],
"abstract": [
"We address the problem of computing the 3-dimensional shape of an arbitrary scene from a set of images taken at known viewpoints. Multi-camera scene reconstruction is a natural generalization of the stereo matching problem. However, it is much more difficult than stereo, primarily due to the difficulty of reasoning about visibility. In this paper, we take an approach that has yielded excellent results for stereo, namely energy minimization via graph cuts. We first give an energy minimization formulation of the multi-camera scene reconstruction problem. The energy that we minimize treats the input images symmetrically, handles visibility properly, and imposes spatial smoothness while preserving discontinuities. As the energy function is NP-hard to minimize exactly, we give a graph cut algorithm that computes a local minimum in a strong sense. We handle all camera configurations where voxel coloring can be used, which is a large and natural class. Experimental data demonstrates the effectiveness of our approach.",
"",
"We introduce an approach to feature-based object recognition, using maximum a posteriori (MAP) estimation under a Markov random field (MRF) model. This approach provides an efficienct solution for a wide class of priors that explicitly model dependencies between individual features of an object. These priors capture phenomena such as the fact that unmatched features due to partial occlusion are generally spatially correlated rather than independent. The main focus of this paper is a special case of the framework that yields a particularly efficient approximation method. We call this special case spatially coherent matching (SCM), as it reflects the spatial correlation among neighboring features of an object. The SCM method operates directly on the image feature map, rather than relying on the graph-based methods used in the general framework. We present some Monte Carlo experiments showing that SCM yields substantial improvements over Hausdorff matching for cluttered scenes and partially occluded objects.",
"In this paper we introduce a new algorithm for image and video texture synthesis. In our approach, patch regions from a sample image or video are transformed and copied to the output and then stitched together along optimal seams to generate a new (and typically larger) output. In contrast to other techniques, the size of the patch is not chosen a-priori, but instead a graph cut technique is used to determine the optimal patch region for any given offset between the input and output texture. Unlike dynamic programming, our graph cut technique for seam optimization is applicable in any dimension. We specifically explore it in 2D and 3D to perform video texture synthesis in addition to regular image synthesis. We present approximative offset search techniques that work well in conjunction with the presented patch size optimization. We show results for synthesizing regular, random, and natural images and videos. We also demonstrate how this method can be used to interactively merge different images to generate new scenes.",
"We present a new image segmentation algorithm based on graph cuts. Our main tool is separation of each pixel p from a special point outside the image by a cut of a minimum cost. Such a cut creates a group of pixels C sub p around each pixel. We show that these groups C sub p are either disjoint or nested in each other and so they give a natural segmentation of the image. In addition this property allows an efficient implementation of the algorithms because for most pixels p the computation of C sub p is not performed on the whole graph. We inspect all C sub p and discard those which are not interesting, for example if they are too small. This procedure automatically groups small components together or merges them into nearby large clusters. Effectively, our segmentation is performed by extracting significant non-intersecting closed contours. We present interesting segmentation results on real and artificial images.",
"This paper describes a new algorithm for solving the N-camera stereo correspondence problem by transforming it into a maximum-flow problem. Once solved, the minimum-cut associated to the maximum-flow yields a disparity surface for the whole image at once. This global approach to stereo analysis provides a more accurate and coherent depth map than the traditional line-by-line stereo. Moreover, the optimality of the depth surface is guaranteed and can be shown to be a generalization of the dynamic programming approach that is widely used in standard stereo. Results show improved depth estimation as well as better handling of depth discontinuities. While the worst case running time is O(n sup 2 d sup 2 log(nd)), the observed average running time is O(n sup 1.2 d sup 1.3 ) for an image size of n pixels and depth resolution d."
]
} |
1904.04297 | 2939498466 | This paper proposes an approach to learn generic multi-modal mesh surface representations using a novel scheme for fusing texture and geometric data. Our approach defines an inverse mapping between different geometric descriptors computed on the mesh surface or its down-sampled version, and the corresponding 2D texture image of the mesh, allowing the construction of fused geometrically augmented images (FGAI). This new fused modality enables us to learn feature representations from 3D data in a highly efficient manner by simply employing standard convolutional neural networks in a transfer-learning mode. In contrast to existing methods, the proposed approach is both computationally and memory efficient, preserves intrinsic geometric information and learns highly discriminative feature representation by effectively fusing shape and texture information at data level. The efficacy of our approach is demonstrated for the tasks of facial action unit detection and expression classification. The extensive experiments conducted on the Bosphorus and BU-4DFE datasets, show that our method produces a significant boost in the performance when compared to state-of-the-art solutions | Most of the research on 3D object classification uses carefully hand-crafted descriptors. @cite_29 presented a comprehensive survey and performance comparison of such descriptors, whereby the authors categorized these traditional descriptors into two groups: descriptors based upon histograms of spatial distributions (, point-clouds). descriptors based upon histograms of local geometric characteristics (, surface normals and curvatures). | {
"cite_N": [
"@cite_29"
],
"mid": [
"2024039087"
],
"abstract": [
"A number of 3D local feature descriptors have been proposed in the literature. It is however, unclear which descriptors are more appropriate for a particular application. A good descriptor should be descriptive, compact, and robust to a set of nuisances. This paper compares ten popular local feature descriptors in the contexts of 3D object recognition, 3D shape retrieval, and 3D modeling. We first evaluate the descriptiveness of these descriptors on eight popular datasets which were acquired using different techniques. We then analyze their compactness using the recall of feature matching per each float value in the descriptor. We also test the robustness of the selected descriptors with respect to support radius variations, Gaussian noise, shot noise, varying mesh resolution, distance to the mesh boundary, keypoint localization error, occlusion, clutter, and dataset size. Moreover, we present the performance results of these descriptors when combined with different 3D keypoint detection methods. We finally analyze the computational efficiency for generating each descriptor."
]
} |
1904.04297 | 2939498466 | This paper proposes an approach to learn generic multi-modal mesh surface representations using a novel scheme for fusing texture and geometric data. Our approach defines an inverse mapping between different geometric descriptors computed on the mesh surface or its down-sampled version, and the corresponding 2D texture image of the mesh, allowing the construction of fused geometrically augmented images (FGAI). This new fused modality enables us to learn feature representations from 3D data in a highly efficient manner by simply employing standard convolutional neural networks in a transfer-learning mode. In contrast to existing methods, the proposed approach is both computationally and memory efficient, preserves intrinsic geometric information and learns highly discriminative feature representation by effectively fusing shape and texture information at data level. The efficacy of our approach is demonstrated for the tasks of facial action unit detection and expression classification. The extensive experiments conducted on the Bosphorus and BU-4DFE datasets, show that our method produces a significant boost in the performance when compared to state-of-the-art solutions | In contrast to learned feature representations, a significant limitation of the traditional hand-crafted descriptors is their lack of generalization across different domains and tasks. These hand-crafted descriptors have been recently outperformed by features learned from raw data using deep CNN. Feature learning capabilities of CNNs from RGB images have been demonstrated in a number of challenging Computer Vision tasks such as object classification and detection, scene recognition, texture classification and image segmentation @cite_0 @cite_4 . Features learned on a generic large scale dataset such as ImageNet have been shown to generalize well across other tasks, , fine-grained classification and scene understanding @cite_32 . Due to their excellent generalization capabilities and impressive performance gain, learned feature representations are believed to be superior compared with traditional hand-crafted features. Despite their success on 2D images, research on learning features from 3D geometric data is still in its infancy compared with its 2D counterpart. Below, we provide an overview of existing 3D deep learning methods. | {
"cite_N": [
"@cite_0",
"@cite_4",
"@cite_32"
],
"mid": [
"2163605009",
"2963037989",
"2062118960"
],
"abstract": [
"We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5 and 17.0 which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overriding in the fully-connected layers we employed a recently-developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3 , compared to 26.2 achieved by the second-best entry.",
"We present YOLO, a new approach to object detection. Prior work on object detection repurposes classifiers to perform detection. Instead, we frame object detection as a regression problem to spatially separated bounding boxes and associated class probabilities. A single neural network predicts bounding boxes and class probabilities directly from full images in one evaluation. Since the whole detection pipeline is a single network, it can be optimized end-to-end directly on detection performance. Our unified architecture is extremely fast. Our base YOLO model processes images in real-time at 45 frames per second. A smaller version of the network, Fast YOLO, processes an astounding 155 frames per second while still achieving double the mAP of other real-time detectors. Compared to state-of-the-art detection systems, YOLO makes more localization errors but is less likely to predict false positives on background. Finally, YOLO learns very general representations of objects. It outperforms other detection methods, including DPM and R-CNN, when generalizing from natural images to other domains like artwork.",
"Recent results indicate that the generic descriptors extracted from the convolutional neural networks are very powerful. This paper adds to the mounting evidence that this is indeed the case. We report on a series of experiments conducted for different recognition tasks using the publicly available code and model of the OverFeat network which was trained to perform object classification on ILSVRC13. We use features extracted from the OverFeat network as a generic image representation to tackle the diverse range of recognition tasks of object image classification, scene recognition, fine grained recognition, attribute detection and image retrieval applied to a diverse set of datasets. We selected these tasks and datasets as they gradually move further away from the original task and data the OverFeat network was trained to solve. Astonishingly, we report consistent superior results compared to the highly tuned state-of-the-art systems in all the visual classification tasks on various datasets. For instance retrieval it consistently outperforms low memory footprint methods except for sculptures dataset. The results are achieved using a linear SVM classifier (or L2 distance in case of retrieval) applied to a feature representation of size 4096 extracted from a layer in the net. The representations are further modified using simple augmentation techniques e.g. jittering. The results strongly suggest that features obtained from deep learning with convolutional nets should be the primary candidate in most visual recognition tasks."
]
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.