aid stringlengths 9 15 | mid stringlengths 7 10 | abstract stringlengths 78 2.56k | related_work stringlengths 92 1.77k | ref_abstract dict |
|---|---|---|---|---|
1704.00939 | 2607065675 | In this paper, we describe a methodology to infer Bullish or Bearish sentiment towards companies brands. More specifically, our approach leverages affective lexica and word embeddings in combination with convolutional neural networks to infer the sentiment of financial news headlines towards a target company. Such architecture was used and evaluated in the context of the SemEval 2017 challenge (task 5, subtask 2), in which it obtained the best performance. | In this work, we exploit the DepecheMood affective lexicon proposed by @cite_13 : this resource has been built in a completely unsupervised fashion, from affective scores assigned by readers to news articles; notably, due to its automated crowd-sourcing-based approach, DepecheMood allows for both high-coverage and high-precision. DepecheMood provides scores for more than 37k entries, on the following affective dimensions: , , , , , , , , . We refer the reader to @cite_13 @cite_19 for more details. | {
"cite_N": [
"@cite_19",
"@cite_13"
],
"mid": [
"2949385208",
"2136961259"
],
"abstract": [
"This article provides a comprehensive investigation on the relations between virality of news articles and the emotions they are found to evoke. Virality, in our view, is a phenomenon with many facets, i.e. under this generic term several different effects of persuasive communication are comprised. By exploiting a high-coverage and bilingual corpus of documents containing metrics of their spread on social networks as well as a massive affective annotation provided by readers, we present a thorough analysis of the interplay between evoked emotions and viral facets. We highlight and discuss our findings in light of a cross-lingual approach: while we discover differences in evoked emotions and corresponding viral effects, we provide preliminary evidence of a generalized explanatory model rooted in the deep structure of emotions: the Valence-Arousal-Dominance (VAD) circumplex. We find that viral facets appear to be consistently affected by particular VAD configurations, and these configurations indicate a clear connection with distinct phenomena underlying persuasive communication.",
"While many lexica annotated with words polarity are available for sentiment analysis, very few tackle the harder task of emotion analysis and are usually quite limited in coverage. In this paper, we present a novel approach for extracting - in a totally automated way - a high-coverage and high-precision lexicon of roughly 37 thousand terms annotated with emotion scores, called DepecheMood. Our approach exploits in an original way 'crowd-sourced' affective annotation implicitly provided by readers of news articles from rappler.com. By providing new state-of-the-art performances in unsupervised settings for regression and classification tasks, even using a na \" ve approach, our experiments show the beneficial impact of harvesting social media data for affective lexicon building."
]
} |
1704.00939 | 2607065675 | In this paper, we describe a methodology to infer Bullish or Bearish sentiment towards companies brands. More specifically, our approach leverages affective lexica and word embeddings in combination with convolutional neural networks to infer the sentiment of financial news headlines towards a target company. Such architecture was used and evaluated in the context of the SemEval 2017 challenge (task 5, subtask 2), in which it obtained the best performance. | A modification of @cite_23 was proposed by Kim @cite_6 for sentence classification, showing how a simple model together with pre-trained word representations can be highly performing. Our method builds on this conv-net method. Further, we took advantage of the rule-based sentiment analyser VADER @cite_20 (for Valence Aware Dictionary for sEntiment Reasoning), which builds upon a sentiment lexicon and a predefined set of simple rules. | {
"cite_N": [
"@cite_20",
"@cite_6",
"@cite_23"
],
"mid": [
"2099813784",
"",
"2952230511"
],
"abstract": [
"The inherent nature of social media content poses serious challenges to practical applications of sentiment analysis. We present VADER, a simple rule-based model for general sentiment analysis, and compare its effectiveness to eleven typical state-of-practice benchmarks including LIWC, ANEW, the General Inquirer, SentiWordNet, and machine learning oriented techniques relying on Naive Bayes, Maximum Entropy, and Support Vector Machine (SVM) algorithms. Using a combination of qualitative and quantitative methods, we first construct and empirically validate a goldstandard list of lexical features (along with their associated sentiment intensity measures) which are specifically attuned to sentiment in microblog-like contexts. We then combine these lexical features with consideration for five general rules that embody grammatical and syntactical conventions for expressing and emphasizing sentiment intensity. Interestingly, using our parsimonious rule-based model to assess the sentiment of tweets, we find that VADER outperforms individual human raters (F1 Classification Accuracy = 0.96 and 0.84, respectively), and generalizes more favorably across contexts than any of our benchmarks.",
"",
"We propose a unified neural network architecture and learning algorithm that can be applied to various natural language processing tasks including: part-of-speech tagging, chunking, named entity recognition, and semantic role labeling. This versatility is achieved by trying to avoid task-specific engineering and therefore disregarding a lot of prior knowledge. Instead of exploiting man-made input features carefully optimized for each task, our system learns internal representations on the basis of vast amounts of mostly unlabeled training data. This work is then used as a basis for building a freely available tagging system with good performance and minimal computational requirements."
]
} |
1704.01006 | 2606516936 | The introduction of automated vehicles without permanent human supervision demands a functional system description, including functional system boundaries and a comprehensive safety analysis. These inputs to the technical development can be identified and analyzed by a scenario-based approach. Furthermore, to establish an economical test and release process, a large number of scenarios must be identified to obtain meaningful test results. Experts are doing well to identify scenarios that are difficult to handle or unlikely to happen. However, experts are unlikely to identify all scenarios possible based on the knowledge they have on hand. Expert knowledge modeled for computer aided processing may help for the purpose of providing a wide range of scenarios. This contribution reviews ontologies as knowledge-based systems in the field of automated vehicles, and proposes a generation of traffic scenes in natural language as a basis for a scenario creation. | @cite_22 propose an environmental model derived from a knowledge base with hierarchical classes and relations between the entities. The ontology is implemented in C++ to provide an environmental model, which was updated by sensor data and used to make online decisions in the test vehicle of the project Stadtpilot @cite_21 . | {
"cite_N": [
"@cite_21",
"@cite_22"
],
"mid": [
"2083016385",
"2144257846"
],
"abstract": [
"The Stadtpilot project aims at autonomous driving on Braunschweig's inner city ring road. For this purpose, an autonomous vehicle called “Leonie” has been developed. In October 2010, after two years of research, “Leonie's” abilities were presented in a public demonstration. This vehicle is one of the first worldwide to show the ability of driving autonomously in real urban traffic scenarios. This paper describes the legal issues and the homologation process for driving autonomously in public traffic in Braunschweig, Germany. It also dwells on the Safety Concept, the system architecture and current research activities.",
"The Stadtpilot project aims at fully automated driving on Braunschweig's inner city ring road. The TU Braunschweig's research vehicle Leonie is one of the first vehicles having the ability of fully automated driving in real urban traffic scenarios. In this paper, we present our approaches for context representation and environment modeling for automated driving. The demonstrated approach allows to provide a simple and yet universal information storage layer for the development of complex driving applications. Moreover, we present our approach for aggregating and fusing information between dynamic traffic objects detected by the sensor systems and a-priori map information."
]
} |
1704.01006 | 2606516936 | The introduction of automated vehicles without permanent human supervision demands a functional system description, including functional system boundaries and a comprehensive safety analysis. These inputs to the technical development can be identified and analyzed by a scenario-based approach. Furthermore, to establish an economical test and release process, a large number of scenarios must be identified to obtain meaningful test results. Experts are doing well to identify scenarios that are difficult to handle or unlikely to happen. However, experts are unlikely to identify all scenarios possible based on the knowledge they have on hand. Expert knowledge modeled for computer aided processing may help for the purpose of providing a wide range of scenarios. This contribution reviews ontologies as knowledge-based systems in the field of automated vehicles, and proposes a generation of traffic scenes in natural language as a basis for a scenario creation. | Xiong @cite_24 proposes a framework for for simulation of test cases based on driving scenarios for automated vehicles. Therefore, the framework consists of an ontology for scenario orchestration (OSO), virtual driver(s), a collection of supporting modules in a scenario management module (SMM), and a scenario observer. One of the main concepts is a simulation supervision linked with the driver models, which calculate all interactions between the entities in the simulation and the necessary tools according to the defined scenario depending on the simulation framework. This framework has been evaluated in two simulation platforms by executing multiple predefined scenarios. | {
"cite_N": [
"@cite_24"
],
"mid": [
"1507161226"
],
"abstract": [
"A scenario in a driving simulator covers what the human participants experience and what the researchers need: the physical scene, predefined traffic flow, simulated vehicles' interactions with the participants and measurements to be collected. Current methodologies used to orchestrate scenarios regarding the interactions have the following drawbacks: 1) Action sequences that simulated vehicles should follow in scenarios are specified without the contexts of each Action; 2) programming languages always include platform-dependent details and are not suitable for context modelling and scenario sharing and 3) there is no mechanism to handle scenarios dynamically and deal with failures to deploy a scenario. To overcome these problems, a concept named Assignment, which represents the task(s) of Virtual Drivers, was first developed to encode the contextual information of proposed Actions for interaction generation, e.g., potential simulated vehicles involved. The Ontology for Scenario Orchestration (OSO) was then developed to model concepts and their relationships in the domain of scenario orchestration including the concept Assignment. It can also provide a file for machine processing. An algorithm named NAUSEA (autoNomous locAl manoeUvre and Scenario orchEstration based on automated action plAnning) was finally generated to utilise Assignments recorded in OSO. Encoded in the driver model SAIL (Scenario-Aware drIver modeL), NAUSEA can be used by a Virtual Driver to control simulated vehicles dynamically. Failed interactions, generated by corresponding Assignments, can be regenerated if necessary. A framework SOAV (Scenario Orchestration with Autonomous simulated Vehicles) was formed to support SAIL - NAUSEA and orchestrate scenarios with autonomous vehicles. Three verification experiments were carried out and showed that SOAV was working properly by producing corresponding interactions based on SAIL NAUSEA and Assignments. They also demonstrated that OSO can provide contextual information in a human-readable and machine processable manner. The OSO evaluation showed that OSO has several advantages such as being readable, flexible etc., but how it can be presented to varieties of audiences needs further examination."
]
} |
1704.01168 | 2605606461 | Informative Bayesian priors are often difficult to elicit, and when this is the case, modelers usually turn to noninformative or objective priors. However, objective priors such as the Jeffreys and reference priors are not tractable to derive for many models of interest. We address this issue by proposing techniques for learning reference prior approximations: we select a parametric family and optimize a black-box lower bound on the reference prior objective to find the member of the family that serves as a good approximation. We experimentally demonstrate the method's effectiveness by recovering Jeffreys priors and learning the Variational Autoencoder's reference prior. | Lafferty & Wasserman (2001) proposed a Markov Chain Monte Carlo (MCMC) method for sampling from a RP. Their approach involves running the Metropolis-Hastings algorithm on the following ratio @cite_14 : where @math is the iteration index, @math is the entropy of the likelihood function, and @math where @math are the parameter samples collected during the previous iteration. | {
"cite_N": [
"@cite_14"
],
"mid": [
"1601253887"
],
"abstract": [
"We present an iterative Markov chain Monte Carlo algorithm for computing reference priors and minimax risk for general parametric families. Our approach uses MCMC techniques based on the Blahut-Arimoto algorithm for computing channel capacity in information theory. We give a statistical analysis of the algorithm, bounding the number of samples required for the stochastic algorithm to closely approximate the deterministic algorithm in each iteration. Simulations are presented for several examples from exponential families. Although we focus on applications to reference priors and minimax risk, the methods and analysis we develop are applicable to a much broader class of optimization problems and iterative algorithms."
]
} |
1704.01168 | 2605606461 | Informative Bayesian priors are often difficult to elicit, and when this is the case, modelers usually turn to noninformative or objective priors. However, objective priors such as the Jeffreys and reference priors are not tractable to derive for many models of interest. We address this issue by proposing techniques for learning reference prior approximations: we select a parametric family and optimize a black-box lower bound on the reference prior objective to find the member of the family that serves as a good approximation. We experimentally demonstrate the method's effectiveness by recovering Jeffreys priors and learning the Variational Autoencoder's reference prior. | The third approach, and the only other that we are aware of for finding approximate RPs, is the (RDM) proposed by (2015). This method focuses on finding a joint RP by minimizing the divergence between a parametric family and the marginal RPs @cite_6 . Since we are concerned with models for which even the marginal RPs are intractable, the RDM is not a relevant point for comparison. | {
"cite_N": [
"@cite_6"
],
"mid": [
"2159666858"
],
"abstract": [
"In multi-parameter models, reference priors typically depend on the parameter or quantity of interest, and it is well known that this is necessary to produce objective posterior distributions with optimal properties. There are, however, many situations where one is simultaneously interested in all the parameters of the model or, more realistically, in functions of them that include aspects such as prediction, and it would then be useful to have a single objective prior that could safely be used to produce reasonable posterior inferences for all the quantities of interest. In this paper, we consider three methods for selecting a single objective prior and study, in a variety of problems including the multinomial problem, whether or not the resulting prior is a reasonable overall prior."
]
} |
1704.00849 | 2608338293 | Building a voice conversion (VC) system from non-parallel speech corpora is challenging but highly valuable in real application scenarios. In most situations, the source and the target speakers do not repeat the same texts or they may even speak different languages. In this case, one possible, although indirect, solution is to build a generative model for speech. Generative models focus on explaining the observations with latent variables instead of learning a pairwise transformation function, thereby bypassing the requirement of speech frame alignment. In this paper, we propose a non-parallel VC framework with a variational autoencoding Wasserstein generative adversarial network (VAW-GAN) that explicitly considers a VC objective when building the speech model. Experimental results corroborate the capability of our framework for building a VC system from unaligned data, and demonstrate improved conversion quality. | To handle VC, many researchers resort to frame-based, segment-based, or cluster-based alignment schemes. One of the most intuitive ways is to apply an automatic speech recognition (ASR) module to the utterances, and proceed with explicit alignment or model adaptation @cite_15 @cite_7 . The ASR module provides every frame with a phonetic label (usually the phonemic states). It is particularly suitable for text-to-speech (TTS) systems because they can readily utilize these labeled frames @cite_17 . A shortcoming with these approaches is that they require an extra mapping to realize cross-lingual VC. To this end, the INCA-based algorithms @cite_10 @cite_16 were proposed to iteratively seek frame-wise correspondence using converted surrogate frames. Another attempt is to separately build frame clusters for the source and the target, and then set up a mapping between them @cite_8 . | {
"cite_N": [
"@cite_7",
"@cite_8",
"@cite_15",
"@cite_16",
"@cite_10",
"@cite_17"
],
"mid": [
"",
"1520370180",
"2290946177",
"",
"2161476805",
"2067175291"
],
"abstract": [
"",
"So far, all conventional voice conversion approaches are text-dependent, i.e., they need equivalent training utterances of source and target speaker. Since several recently proposed applications call for renouncing this requirement, in this paper, we present an algorithm which finds corresponding time frames within text-independent training data. The performance of this algorithm is tested by means of a voice conversion framework based on linear transformation of the spectral envelope. Experimental results are reported on a Spanish cross-gender corpus utilizing several objective error measures.",
"To convert one speaker's voice to another's, the mapping of the corresponding speech segments from source speaker to target speaker must be obtained first. In parallel voice conversion, normally dynamic time warping (DTW) method is used to align signals of source and target voices. However, for conversion between non-parallel speech data, the DTW based mapping method does not work. In this paper, we propose to use a DNN-HMM recognizer to recognize each frame for both source and target speech signals. The vector of pseudo likelihood is then used to represent the frame. Similarity between two frames is measured with the distance between the vectors. A clustering method is used to group both source and target frames. Frame mapping from source to target is then established based on the clustering result. The experiments show that the proposed method can generate similar conversion results compared to parallel voice conversion.",
"",
"Most existing voice conversion systems, particularly those based on Gaussian mixture models, require a set of paired acoustic vectors from the source and target speakers to learn their corresponding transformation function. The alignment of phonetically equivalent source and target vectors is not problematic when the training corpus is parallel, which means that both speakers utter the same training sentences. However, in some practical situations, such as cross-lingual voice conversion, it is not possible to obtain such parallel utterances. With an aim towards increasing the versatility of current voice conversion systems, this paper proposes a new iterative alignment method that allows pairing phonetically equivalent acoustic vectors from nonparallel utterances from different speakers, even under cross-lingual conditions. This method is based on existing voice conversion techniques, and it does not require any phonetic or linguistic information. Subjective evaluation experiments show that the performance of the resulting voice conversion system is very similar to that of an equivalent system trained on a parallel corpus.",
"In this paper, we propose a simple and efficient non-parallel training scheme for voice conversion (VC). First, the speaker models are adapted from the background model using maximum a posteriori (MAP) technique. Then, by utilizing the parameters of adapted speaker models, the Gaussian normalization and mean transformation methods are proposed for VC, respectively. In addition, to improve the conversion performance of the proposed methods, a combination approach is further presented. Finally, objective and subjective experiments are carried out to evaluate the performance of the proposed scheme, the results demonstrate that our scheme can obtain comparable performance with the traditional GMM method based on parallel corpus."
]
} |
1704.01133 | 2952325303 | We propose a vision-based method that localizes a ground vehicle using publicly available satellite imagery as the only prior knowledge of the environment. Our approach takes as input a sequence of ground-level images acquired by the vehicle as it navigates, and outputs an estimate of the vehicle's pose relative to a georeferenced satellite image. We overcome the significant viewpoint and appearance variations between the images through a neural multi-view model that learns location-discriminative embeddings in which ground-level images are matched with their corresponding satellite view of the scene. We use this learned function as an observation model in a filtering framework to maintain a distribution over the vehicle's pose. We evaluate our method on different benchmark datasets and demonstrate its ability localize ground-level images in environments novel relative to training, despite the challenges of significant viewpoint and appearance variations. | Single-view approaches assume access to reference databases that consist of geotagged images of the target environment acquired from vantage points similar to that of the query image (i.e., other ground-level images). These databases may come in the form of collections of Internet-based geotagged images, such as those available via photo sharing websites or Google Street View , or maintained in maps of the environment (e.g., previously recorded using GPS information or generated via SLAM) . The primary challenges to visual place recognition arise due to variations in viewpoint, variations in appearance that result from changes in environment structure, illumination, and seasons, as well as to perceptual aliasing. Much of the early work attempts to mitigate some of these challenges by using hand-crafted features that exhibit some robustness to transformations in scale and rotation, as well as to slight variations in illumination (e.g., SIFT @cite_1 and SURF @cite_7 ), or a combination of visual and textual (i.e., image tags) features . Place recognition then follows as image retrieval, i.e., image-to-image matching-based search against the database . | {
"cite_N": [
"@cite_1",
"@cite_7"
],
"mid": [
"2151103935",
"1677409904"
],
"abstract": [
"This paper presents a method for extracting distinctive invariant features from images that can be used to perform reliable matching between different views of an object or scene. The features are invariant to image scale and rotation, and are shown to provide robust matching across a substantial range of affine distortion, change in 3D viewpoint, addition of noise, and change in illumination. The features are highly distinctive, in the sense that a single feature can be correctly matched with high probability against a large database of features from many images. This paper also describes an approach to using these features for object recognition. The recognition proceeds by matching individual features to a database of features from known objects using a fast nearest-neighbor algorithm, followed by a Hough transform to identify clusters belonging to a single object, and finally performing verification through least-squares solution for consistent pose parameters. This approach to recognition can robustly identify objects among clutter and occlusion while achieving near real-time performance.",
"In this paper, we present a novel scale- and rotation-invariant interest point detector and descriptor, coined SURF (Speeded Up Robust Features). It approximates or even outperforms previously proposed schemes with respect to repeatability, distinctiveness, and robustness, yet can be computed and compared much faster. This is achieved by relying on integral images for image convolutions; by building on the strengths of the leading existing detectors and descriptors (in casu, using a Hessian matrix-based measure for the detector, and a distribution-based descriptor); and by simplifying these methods to the essential. This leads to a combination of novel detection, description, and matching steps. The paper presents experimental results on a standard evaluation set, as well as on imagery obtained in the context of a real-life object recognition application. Both show SURF's strong performance."
]
} |
1704.01069 | 2783119789 | We introduce Multi-Expert Region-based CNN (ME R-CNN) which is equipped with multiple experts and built on top of the R-CNN framework known to be one of the state-of-the-art object detection methods. ME R-CNN focuses in better capturing the appearance variations caused by different shapes, poses, and viewing angles. The proposed approach consists of three experts each responsible for objects with particular shapes: horizontally elongated, square-like, and vertically elongated. On top of using selective search which provides a compact, yet effective set of region of interests (RoIs) for object detection, we augmented the set by also employing the exhaustive search for training only. Incorporating the exhaustive search can provide complementary advantages: i) it captures the multitude of neighboring RoIs missed by the selective search, and thus ii) provide significantly larger amount of training examples. We show that the ME R-CNN architecture provides considerable performance increase over the baselines on PASCAL VOC 07, 12, and MS COCO datasets. | Object Detection. Object detection is one of the most challenging tasks in computer vision. Prior to the introduction of CNNs, non-CNN based object detection approaches, such as HOG-SVM, DPM, etc., were widely used for classifying RoIs into corresponding object categories @cite_36 @cite_33 @cite_14 @cite_25 . Within the past several years, multiple attempts have been made to use CNNs for object detection. Prominent methods among them are R-CNN @cite_21 and its descendants @cite_2 @cite_3 @cite_35 @cite_12 @cite_26 that provided the state-of-the-art performance. | {
"cite_N": [
"@cite_35",
"@cite_14",
"@cite_26",
"@cite_33",
"@cite_36",
"@cite_21",
"@cite_3",
"@cite_2",
"@cite_25",
"@cite_12"
],
"mid": [
"",
"2066477856",
"2953106684",
"2168356304",
"2161969291",
"2102605133",
"",
"2950800384",
"1989684337",
"2179352600"
],
"abstract": [
"",
"State-of-the-art object detectors typically use shape information as a low level feature representation to capture the local structure of an object. This paper shows that early fusion of shape and color, as is popular in image classification, leads to a significant drop in performance for object detection. Moreover, such approaches also yields suboptimal results for object categories with varying importance of color and shape. In this paper we propose the use of color attributes as an explicit color representation for object detection. Color attributes are compact, computationally efficient, and when combined with traditional shape features provide state-of-the-art results for object detection. Our method is tested on the PASCAL VOC 2007 and 2009 datasets and results clearly show that our method improves over state-of-the-art techniques despite its simplicity. We also introduce a new dataset consisting of cartoon character images in which color plays a pivotal role. On this dataset, our approach yields a significant gain of 14 in mean AP over conventional state-of-the-art methods.",
"State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN 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. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features---using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model, 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, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks. Code has been made publicly available.",
"We describe an object detection system based on mixtures of multiscale deformable part models. Our system is able to represent highly variable object classes and achieves state-of-the-art results in the PASCAL object detection challenges. While deformable part models have become quite popular, their value had not been demonstrated on difficult benchmarks such as the PASCAL data sets. Our system relies on new methods for discriminative training with partially labeled data. We combine a margin-sensitive approach for data-mining hard negative examples with a formalism we call latent SVM. A latent SVM is a reformulation of MI--SVM in terms of latent variables. A latent SVM is semiconvex, and the training problem becomes convex once latent information is specified for the positive examples. This leads to an iterative training algorithm that alternates between fixing latent values for positive examples and optimizing the latent SVM objective function.",
"We study the question of feature sets for robust visual object recognition; adopting linear SVM based human detection as a test case. After reviewing existing edge and gradient based descriptors, we show experimentally that grids of histograms of oriented gradient (HOG) descriptors significantly outperform existing feature sets for human detection. We study the influence of each stage of the computation on performance, concluding that fine-scale gradients, fine orientation binning, relatively coarse spatial binning, and high-quality local contrast normalization in overlapping descriptor blocks are all important for good results. The new approach gives near-perfect separation on the original MIT pedestrian database, so we introduce a more challenging dataset containing over 1800 annotated human images with a large range of pose variations and backgrounds.",
"Object detection performance, as measured on the canonical PASCAL VOC dataset, has plateaued in the last few years. The best-performing methods are complex ensemble systems that typically combine multiple low-level image features with high-level context. In this paper, we propose a simple and scalable detection algorithm that improves mean average precision (mAP) by more than 30 relative to the previous best result on VOC 2012 -- achieving a mAP of 53.3 . Our approach combines two key insights: (1) one can apply high-capacity convolutional neural networks (CNNs) to bottom-up region proposals in order to localize and segment objects and (2) when labeled training data is scarce, supervised pre-training for an auxiliary task, followed by domain-specific fine-tuning, yields a significant performance boost. Since we combine region proposals with CNNs, we call our method R-CNN: Regions with CNN features. We also present experiments that provide insight into what the network learns, revealing a rich hierarchy of image features. Source code for the complete system is available at http: www.cs.berkeley.edu rbg rcnn.",
"",
"We present region-based, fully convolutional networks for accurate and efficient object detection. In contrast to previous region-based detectors such as Fast Faster R-CNN that apply a costly per-region subnetwork hundreds of times, our region-based detector is fully convolutional with almost all computation shared on the entire image. To achieve this goal, we propose position-sensitive score maps to address a dilemma between translation-invariance in image classification and translation-variance in object detection. Our method can thus naturally adopt fully convolutional image classifier backbones, such as the latest Residual Networks (ResNets), for object detection. We show competitive results on the PASCAL VOC datasets (e.g., 83.6 mAP on the 2007 set) with the 101-layer ResNet. Meanwhile, our result is achieved at a test-time speed of 170ms per image, 2.5-20x faster than the Faster R-CNN counterpart. Code is made publicly available at: this https URL",
"This paper proposes a conceptually simple but surprisingly powerful method which combines the effectiveness of a discriminative object detector with the explicit correspondence offered by a nearest-neighbor approach. The method is based on training a separate linear SVM classifier for every exemplar in the training set. Each of these Exemplar-SVMs is thus defined by a single positive instance and millions of negatives. While each detector is quite specific to its exemplar, we empirically observe that an ensemble of such Exemplar-SVMs offers surprisingly good generalization. Our performance on the PASCAL VOC detection task is on par with the much more complex latent part-based model of , at only a modest computational cost increase. But the central benefit of our approach is that it creates an explicit association between each detection and a single training exemplar. Because most detections show good alignment to their associated exemplar, it is possible to transfer any available exemplar meta-data (segmentation, geometric structure, 3D model, etc.) directly onto the detections, which can then be used as part of overall scene understanding.",
"Existing deep convolutional neural networks (CNNs) require a fixed-size (e.g. 224×224) input image. This requirement is “artificial” and may hurt the recognition accuracy for the images or sub-images of an arbitrary size scale. In this work, we equip the networks with a more principled pooling strategy, “spatial pyramid pooling”, to eliminate the above requirement. The new network structure, called SPP-net, can generate a fixed-length representation regardless of image size scale. By removing the fixed-size limitation, we can improve all CNN-based image classification methods in general. Our SPP-net achieves state-of-the-art accuracy on the datasets of ImageNet 2012, Pascal VOC 2007, and Caltech101."
]
} |
1704.01069 | 2783119789 | We introduce Multi-Expert Region-based CNN (ME R-CNN) which is equipped with multiple experts and built on top of the R-CNN framework known to be one of the state-of-the-art object detection methods. ME R-CNN focuses in better capturing the appearance variations caused by different shapes, poses, and viewing angles. The proposed approach consists of three experts each responsible for objects with particular shapes: horizontally elongated, square-like, and vertically elongated. On top of using selective search which provides a compact, yet effective set of region of interests (RoIs) for object detection, we augmented the set by also employing the exhaustive search for training only. Incorporating the exhaustive search can provide complementary advantages: i) it captures the multitude of neighboring RoIs missed by the selective search, and thus ii) provide significantly larger amount of training examples. We show that the ME R-CNN architecture provides considerable performance increase over the baselines on PASCAL VOC 07, 12, and MS COCO datasets. | Although having achieved the top-notch performance, R-CNNs have not yet exploited some of the effective strategies which conventional object detection methods commonly use for boosting the performance. While the R-CNNs rely on heuristics to select hard negative examples, @cite_23 and @cite_6 used the online hard example mining (OHEM) to automatically select hard examples with high optimization loss in every iteration of training. These approaches were motivated by the offline bootstrapping idea for training a classical object detection method @cite_36 . | {
"cite_N": [
"@cite_36",
"@cite_6",
"@cite_23"
],
"mid": [
"2161969291",
"2952815469",
""
],
"abstract": [
"We study the question of feature sets for robust visual object recognition; adopting linear SVM based human detection as a test case. After reviewing existing edge and gradient based descriptors, we show experimentally that grids of histograms of oriented gradient (HOG) descriptors significantly outperform existing feature sets for human detection. We study the influence of each stage of the computation on performance, concluding that fine-scale gradients, fine orientation binning, relatively coarse spatial binning, and high-quality local contrast normalization in overlapping descriptor blocks are all important for good results. The new approach gives near-perfect separation on the original MIT pedestrian database, so we introduce a more challenging dataset containing over 1800 annotated human images with a large range of pose variations and backgrounds.",
"How do we learn an object detector that is invariant to occlusions and deformations? Our current solution is to use a data-driven strategy -- collect large-scale datasets which have object instances under different conditions. The hope is that the final classifier can use these examples to learn invariances. But is it really possible to see all the occlusions in a dataset? We argue that like categories, occlusions and object deformations also follow a long-tail. Some occlusions and deformations are so rare that they hardly happen; yet we want to learn a model invariant to such occurrences. In this paper, we propose an alternative solution. We propose to learn an adversarial network that generates examples with occlusions and deformations. The goal of the adversary is to generate examples that are difficult for the object detector to classify. In our framework both the original detector and adversary are learned in a joint manner. Our experimental results indicate a 2.3 mAP boost on VOC07 and a 2.6 mAP boost on VOC2012 object detection challenge compared to the Fast-RCNN pipeline. We also release the code for this paper.",
""
]
} |
1704.01069 | 2783119789 | We introduce Multi-Expert Region-based CNN (ME R-CNN) which is equipped with multiple experts and built on top of the R-CNN framework known to be one of the state-of-the-art object detection methods. ME R-CNN focuses in better capturing the appearance variations caused by different shapes, poses, and viewing angles. The proposed approach consists of three experts each responsible for objects with particular shapes: horizontally elongated, square-like, and vertically elongated. On top of using selective search which provides a compact, yet effective set of region of interests (RoIs) for object detection, we augmented the set by also employing the exhaustive search for training only. Incorporating the exhaustive search can provide complementary advantages: i) it captures the multitude of neighboring RoIs missed by the selective search, and thus ii) provide significantly larger amount of training examples. We show that the ME R-CNN architecture provides considerable performance increase over the baselines on PASCAL VOC 07, 12, and MS COCO datasets. | Mixture-of-Experts Models. Multiple experts embedded in the proposed ME R-CNN is based on the concept of mixture-of-experts models. The mixture-of-experts model is used to better estimate the probability distribution of a composite data with large variation (e.g., Gaussian mixture model @cite_15 ). In the image domain, object appearances can also show large variations according to their shapes, poses, and viewing angles. @cite_33 nicely illustrates the importance of using a mixture of models by presenting two models, each of which captures the appearance of the front and the side view of a bicycle. Accordingly, many recent approaches @cite_32 @cite_33 @cite_1 have shown that using the mixture-of-experts model for advanced object detection is very effective. However, to date to the best of our knowledge, none of the CNN-based object detection methods have incorporated the mixture-of-experts model into their architectures. | {
"cite_N": [
"@cite_1",
"@cite_15",
"@cite_32",
"@cite_33"
],
"mid": [
"2155511848",
"2142493339",
"",
"2168356304"
],
"abstract": [
"In this paper, we describe a statistical method for 3D object detection. We represent the statistics of both object appearance and \"non-object\" appearance using a product of histograms. Each histogram represents the joint statistics of a subset of wavelet coefficients and their position on the object. Our approach is to use many such histograms representing a wide variety of visual attributes. Using this method, we have developed the first algorithm that can reliably detect human faces with out-of-plane rotation and the first algorithm that can reliably detect passenger cars over a wide range of viewpoints.",
"Detecting moving objects on mobile cameras in real-time is a challenging problem due to the computational limits and the motions of the camera. In this paper, we propose a method for moving object detection on non-stationary cameras running within 5.8 milliseconds (ms) on a PC, and real-time on mobile devices. To achieve real time capability with satisfying performance, the proposed method models the background through dual-mode single Gaussian model (SGM) with age and compensates the motion of the camera by mixing neighboring models. Modeling through dual-mode SGM prevents the background model from being contaminated by foreground pixels, while still allowing the model to be able to adapt to changes of the background. Mixing neighboring models reduces the errors arising from motion compensation and their influences are further reduced by keeping the age of the model. Also, to decrease computation load, the proposed method applies one dual-mode SGM to multiple pixels without performance degradation. Experimental results show the computational lightness and the real-time capability of our method on a smart phone with robust detection performances.",
"",
"We describe an object detection system based on mixtures of multiscale deformable part models. Our system is able to represent highly variable object classes and achieves state-of-the-art results in the PASCAL object detection challenges. While deformable part models have become quite popular, their value had not been demonstrated on difficult benchmarks such as the PASCAL data sets. Our system relies on new methods for discriminative training with partially labeled data. We combine a margin-sensitive approach for data-mining hard negative examples with a formalism we call latent SVM. A latent SVM is a reformulation of MI--SVM in terms of latent variables. A latent SVM is semiconvex, and the training problem becomes convex once latent information is specified for the positive examples. This leads to an iterative training algorithm that alternates between fixing latent values for positive examples and optimizing the latent SVM objective function."
]
} |
1704.01069 | 2783119789 | We introduce Multi-Expert Region-based CNN (ME R-CNN) which is equipped with multiple experts and built on top of the R-CNN framework known to be one of the state-of-the-art object detection methods. ME R-CNN focuses in better capturing the appearance variations caused by different shapes, poses, and viewing angles. The proposed approach consists of three experts each responsible for objects with particular shapes: horizontally elongated, square-like, and vertically elongated. On top of using selective search which provides a compact, yet effective set of region of interests (RoIs) for object detection, we augmented the set by also employing the exhaustive search for training only. Incorporating the exhaustive search can provide complementary advantages: i) it captures the multitude of neighboring RoIs missed by the selective search, and thus ii) provide significantly larger amount of training examples. We show that the ME R-CNN architecture provides considerable performance increase over the baselines on PASCAL VOC 07, 12, and MS COCO datasets. | RoI Generation. One of the conventional ways to generate RoIs is to use multi-scale sliding windows @cite_37 @cite_36 @cite_33 @cite_8 @cite_14 @cite_25 @cite_20 which can be considered as a dense' search. To avoid impractical computational complexity, the search space is confined to a regular grid and a fixed set of scales and aspect ratios. The branch and bound strategy was found to reduce the search space even more by using optimal windows within an image @cite_16 @cite_11 . | {
"cite_N": [
"@cite_37",
"@cite_14",
"@cite_33",
"@cite_8",
"@cite_36",
"@cite_16",
"@cite_25",
"@cite_20",
"@cite_11"
],
"mid": [
"1991367009",
"2066477856",
"2168356304",
"",
"2161969291",
"2129052208",
"1989684337",
"2164598857",
""
],
"abstract": [
"We propose a unified approach for bottom-up hierarchical image segmentation and object candidate generation for recognition, called Multiscale Combinatorial Grouping (MCG). For this purpose, we first develop a fast normalized cuts algorithm. We then propose a high-performance hierarchical segmenter that makes effective use of multiscale information. Finally, we propose a grouping strategy that combines our multiscale regions into highly-accurate object candidates by exploring efficiently their combinatorial space. We conduct extensive experiments on both the BSDS500 and on the PASCAL 2012 segmentation datasets, showing that MCG produces state-of-the-art contours, hierarchical regions and object candidates.",
"State-of-the-art object detectors typically use shape information as a low level feature representation to capture the local structure of an object. This paper shows that early fusion of shape and color, as is popular in image classification, leads to a significant drop in performance for object detection. Moreover, such approaches also yields suboptimal results for object categories with varying importance of color and shape. In this paper we propose the use of color attributes as an explicit color representation for object detection. Color attributes are compact, computationally efficient, and when combined with traditional shape features provide state-of-the-art results for object detection. Our method is tested on the PASCAL VOC 2007 and 2009 datasets and results clearly show that our method improves over state-of-the-art techniques despite its simplicity. We also introduce a new dataset consisting of cartoon character images in which color plays a pivotal role. On this dataset, our approach yields a significant gain of 14 in mean AP over conventional state-of-the-art methods.",
"We describe an object detection system based on mixtures of multiscale deformable part models. Our system is able to represent highly variable object classes and achieves state-of-the-art results in the PASCAL object detection challenges. While deformable part models have become quite popular, their value had not been demonstrated on difficult benchmarks such as the PASCAL data sets. Our system relies on new methods for discriminative training with partially labeled data. We combine a margin-sensitive approach for data-mining hard negative examples with a formalism we call latent SVM. A latent SVM is a reformulation of MI--SVM in terms of latent variables. A latent SVM is semiconvex, and the training problem becomes convex once latent information is specified for the positive examples. This leads to an iterative training algorithm that alternates between fixing latent values for positive examples and optimizing the latent SVM objective function.",
"",
"We study the question of feature sets for robust visual object recognition; adopting linear SVM based human detection as a test case. After reviewing existing edge and gradient based descriptors, we show experimentally that grids of histograms of oriented gradient (HOG) descriptors significantly outperform existing feature sets for human detection. We study the influence of each stage of the computation on performance, concluding that fine-scale gradients, fine orientation binning, relatively coarse spatial binning, and high-quality local contrast normalization in overlapping descriptor blocks are all important for good results. The new approach gives near-perfect separation on the original MIT pedestrian database, so we introduce a more challenging dataset containing over 1800 annotated human images with a large range of pose variations and backgrounds.",
"In this work we use Branch-and-Bound (BB) to efficiently detect objects with deformable part models. Instead of evaluating the classifier score exhaustively over image locations and scales, we use BB to focus on promising image locations. The core problem is to compute bounds that accommodate part deformations; for this we adapt the Dual Trees data structure [7] to our problem. We evaluate our approach using Mixture-of-Deformable Part Models [4]. We obtain exactly the same results but are 10-20 times faster on average. We also develop a multiple-object detection variation of the system, where hypotheses for 20 categories are inserted in a common priority queue. For the problem of finding the strongest category in an image this results in a 100-fold speedup.",
"This paper proposes a conceptually simple but surprisingly powerful method which combines the effectiveness of a discriminative object detector with the explicit correspondence offered by a nearest-neighbor approach. The method is based on training a separate linear SVM classifier for every exemplar in the training set. Each of these Exemplar-SVMs is thus defined by a single positive instance and millions of negatives. While each detector is quite specific to its exemplar, we empirically observe that an ensemble of such Exemplar-SVMs offers surprisingly good generalization. Our performance on the PASCAL VOC detection task is on par with the much more complex latent part-based model of , at only a modest computational cost increase. But the central benefit of our approach is that it creates an explicit association between each detection and a single training exemplar. Because most detections show good alignment to their associated exemplar, it is possible to transfer any available exemplar meta-data (segmentation, geometric structure, 3D model, etc.) directly onto the detections, which can then be used as part of overall scene understanding.",
"This paper describes a machine learning approach for visual object detection which is capable of processing images extremely rapidly and achieving high detection rates. This work is distinguished by three key contributions. The first is the introduction of a new image representation called the \"integral image\" which allows the features used by our detector to be computed very quickly. The second is a learning algorithm, based on AdaBoost, which selects a small number of critical visual features from a larger set and yields extremely efficient classifiers. The third contribution is a method for combining increasingly more complex classifiers in a \"cascade\" which allows background regions of the image to be quickly discarded while spending more computation on promising object-like regions. The cascade can be viewed as an object specific focus-of-attention mechanism which unlike previous approaches provides statistical guarantees that discarded regions are unlikely to contain the object of interest. In the domain of face detection the system yields detection rates comparable to the best previous systems. Used in real-time applications, the detector runs at 15 frames per second without resorting to image differencing or skin color detection.",
""
]
} |
1704.01069 | 2783119789 | We introduce Multi-Expert Region-based CNN (ME R-CNN) which is equipped with multiple experts and built on top of the R-CNN framework known to be one of the state-of-the-art object detection methods. ME R-CNN focuses in better capturing the appearance variations caused by different shapes, poses, and viewing angles. The proposed approach consists of three experts each responsible for objects with particular shapes: horizontally elongated, square-like, and vertically elongated. On top of using selective search which provides a compact, yet effective set of region of interests (RoIs) for object detection, we augmented the set by also employing the exhaustive search for training only. Incorporating the exhaustive search can provide complementary advantages: i) it captures the multitude of neighboring RoIs missed by the selective search, and thus ii) provide significantly larger amount of training examples. We show that the ME R-CNN architecture provides considerable performance increase over the baselines on PASCAL VOC 07, 12, and MS COCO datasets. | Instead of going dense', some methods employed relatively sparse' searching approaches by introducing the concept of objectness. @cite_24 used an objectness quality function to discard sub-search spaces whose objectness scores are under a certain threshold, where object detector becomes an objectness quality function. Instead of using the object detector, @cite_19 introduces a generic objectness measure, to estimate how likely it is for a region to contain object of any category using saliency, color contrast, edge density, and boundary information. Several more approaches @cite_22 @cite_31 @cite_0 @cite_30 @cite_28 @cite_9 to generate RoIs based on objectness characteristics have been introduced afterwards. Recently, @cite_26 introduced a region proposal network (RPN) incorporated into the CNN which also generates RoIs based on the objectness. | {
"cite_N": [
"@cite_30",
"@cite_26",
"@cite_22",
"@cite_28",
"@cite_9",
"@cite_24",
"@cite_19",
"@cite_0",
"@cite_31"
],
"mid": [
"2104446196",
"2953106684",
"2046382188",
"2088049833",
"7746136",
"2106874006",
"2128715914",
"",
""
],
"abstract": [
"Current top performing Pascal VOC object detectors employ detection proposals to guide the search for objects thereby avoiding exhaustive sliding window search across images. Despite the popularity of detection proposals, it is unclear which trade-offs are made when using them during object detection. We provide an in depth analysis of ten object proposal methods along with four baselines regarding ground truth annotation recall (on Pascal VOC 2007 and ImageNet 2013), repeatability, and impact on DPM detector performance. Our findings show common weaknesses of existing methods, and provide insights to choose the most adequate method for different settings.",
"State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN 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. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features---using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model, 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, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks. Code has been made publicly available.",
"We present a novel framework to generate and rank plausible hypotheses for the spatial extent of objects in images using bottom-up computational processes and mid-level selection cues. The object hypotheses are represented as figure-ground segmentations, and are extracted automatically, without prior knowledge of the properties of individual object classes, by solving a sequence of Constrained Parametric Min-Cut problems (CPMC) on a regular image grid. In a subsequent step, we learn to rank the corresponding segments by training a continuous model to predict how likely they are to exhibit real-world regularities (expressed as putative overlap with ground truth) based on their mid-level region properties, then diversify the estimated overlap score using maximum marginal relevance measures. We show that this algorithm significantly outperforms the state of the art for low-level segmentation in the VOC 2009 and 2010 data sets. In our companion papers [1], [2], we show that the algorithm can be used, successfully, in a segmentation-based visual object category recognition pipeline. This architecture ranked first in the VOC2009 and VOC2010 image segmentation and labeling challenges.",
"This paper addresses the problem of generating possible object locations for use in object recognition. We introduce selective search which combines the strength of both an exhaustive search and segmentation. Like segmentation, we use the image structure to guide our sampling process. Like exhaustive search, we aim to capture all possible object locations. Instead of a single technique to generate possible object locations, we diversify our search and use a variety of complementary image partitionings to deal with as many image conditions as possible. Our selective search results in a small set of data-driven, class-independent, high quality locations, yielding 99 recall and a Mean Average Best Overlap of 0.879 at 10,097 locations. The reduced number of locations compared to an exhaustive search enables the use of stronger machine learning techniques and stronger appearance models for object recognition. In this paper we show that our selective search enables the use of the powerful Bag-of-Words model for recognition. The selective search software is made publicly available (Software: http: disi.unitn.it uijlings SelectiveSearch.html ).",
"The use of object proposals is an effective recent approach for increasing the computational efficiency of object detection. We propose a novel method for generating object bounding box proposals using edges. Edges provide a sparse yet informative representation of an image. Our main observation is that the number of contours that are wholly contained in a bounding box is indicative of the likelihood of the box containing an object. We propose a simple box objectness score that measures the number of edges that exist in the box minus those that are members of contours that overlap the box’s boundary. Using efficient data structures, millions of candidate boxes can be evaluated in a fraction of a second, returning a ranked set of a few thousand top-scoring proposals. Using standard metrics, we show results that are significantly more accurate than the current state-of-the-art while being faster to compute. In particular, given just 1000 proposals we achieve over 96 object recall at overlap threshold of 0.5 and over 75 recall at the more challenging overlap of 0.7. Our approach runs in 0.25 seconds and we additionally demonstrate a near real-time variant with only minor loss in accuracy.",
"Most successful object recognition systems rely on binary classification, deciding only if an object is present or not, but not providing information on the actual object location. To estimate the object's location, one can take a sliding window approach, but this strongly increases the computational cost because the classifier or similarity function has to be evaluated over a large set of candidate subwindows. In this paper, we propose a simple yet powerful branch and bound scheme that allows efficient maximization of a large class of quality functions over all possible subimages. It converges to a globally optimal solution typically in linear or even sublinear time, in contrast to the quadratic scaling of exhaustive or sliding window search. We show how our method is applicable to different object detection and image retrieval scenarios. The achieved speedup allows the use of classifiers for localization that formerly were considered too slow for this task, such as SVMs with a spatial pyramid kernel or nearest-neighbor classifiers based on the lambda2 distance. We demonstrate state-of-the-art localization performance of the resulting systems on the UIUC Cars data set, the PASCAL VOC 2006 data set, and in the PASCAL VOC 2007 competition.",
"We present a generic objectness measure, quantifying how likely it is for an image window to contain an object of any class. We explicitly train it to distinguish objects with a well-defined boundary in space, such as cows and telephones, from amorphous background elements, such as grass and road. The measure combines in a Bayesian framework several image cues measuring characteristics of objects, such as appearing different from their surroundings and having a closed boundary. This includes an innovative cue measuring the closed boundary characteristic. In experiments on the challenging PASCAL VOC 07 dataset, we show this new cue to outperform a state-of-the-art saliency measure [17], and the combined measure to perform better than any cue alone. Finally, we show how to sample windows from an image according to their objectness distribution and give an algorithm to employ them as location priors for modern class-specific object detectors. In experiments on PASCAL VOC 07 we show this greatly reduces the number of windows evaluated by class-specific object detectors.",
"",
""
]
} |
1704.01069 | 2783119789 | We introduce Multi-Expert Region-based CNN (ME R-CNN) which is equipped with multiple experts and built on top of the R-CNN framework known to be one of the state-of-the-art object detection methods. ME R-CNN focuses in better capturing the appearance variations caused by different shapes, poses, and viewing angles. The proposed approach consists of three experts each responsible for objects with particular shapes: horizontally elongated, square-like, and vertically elongated. On top of using selective search which provides a compact, yet effective set of region of interests (RoIs) for object detection, we augmented the set by also employing the exhaustive search for training only. Incorporating the exhaustive search can provide complementary advantages: i) it captures the multitude of neighboring RoIs missed by the selective search, and thus ii) provide significantly larger amount of training examples. We show that the ME R-CNN architecture provides considerable performance increase over the baselines on PASCAL VOC 07, 12, and MS COCO datasets. | Going Wider with CNN. One of the major innovations introduced into ME R-CNN is that the network has expanded in width, where the network width refers to the number of nodes in each layer. This is to equip the network with multiple number of specialized experts to better capture variations of object appearance. There have already been several attempts where the width of CNN architecture was expanded. @cite_18 splits each layer into two parallel layers in order to fully use two GPUs in a parallel fashion. @cite_5 uses the inception module which employs multiple parallel layers in order to make use of dense sets of different sized convolutional filters. Several other approaches @cite_27 @cite_7 @cite_10 also introduced widened networks for the task of co-learning multiple tasks in a single framework. | {
"cite_N": [
"@cite_18",
"@cite_7",
"@cite_27",
"@cite_5",
"@cite_10"
],
"mid": [
"2543366260",
"2951825824",
"2949295283",
"2950179405",
"2963498646"
],
"abstract": [
"In this paper, we introduce a novel fusion method that can enhance object detection performance by fusing decisions from two different types of computer vision tasks: object detection and image classification. In the proposed work, the class label of an image obtained from image classification is viewed as prior knowledge about existence or non-existence of certain objects. The prior knowledge is then fused with the decisions of object detection to improve detection accuracy by mitigating false positives of an object detector that are strongly contradicted with the prior knowledge. A recently introduced novel fusion approach called dynamic belief fusion (DBF) is used to fuse the detector output with the classification prior. Experimental results show that the detection performance of all the detection algorithms used in the proposed work is improved on benchmark datasets via the proposed fusion framework.",
"Many previous methods have showed the importance of considering semantically relevant objects for performing event recognition, yet none of the methods have exploited the power of deep convolutional neural networks to directly integrate relevant object information into a unified network. We present a novel unified deep CNN architecture which integrates architecturally different, yet semantically-related object detection networks to enhance the performance of the event recognition task. Our architecture allows the sharing of the convolutional layers and a fully connected layer which effectively integrates event recognition, rigid object detection and non-rigid object detection.",
"Semantic segmentation research has recently witnessed rapid progress, but many leading methods are unable to identify object instances. In this paper, we present Multi-task Network Cascades for instance-aware semantic segmentation. Our model consists of three networks, respectively differentiating instances, estimating masks, and categorizing objects. These networks form a cascaded structure, and are designed to share their convolutional features. We develop an algorithm for the nontrivial end-to-end training of this causal, cascaded structure. Our solution is a clean, single-step training framework and can be generalized to cascades that have more stages. We demonstrate state-of-the-art instance-aware semantic segmentation accuracy on PASCAL VOC. Meanwhile, our method takes only 360ms testing an image using VGG-16, which is two orders of magnitude faster than previous systems for this challenging problem. As a by product, our method also achieves compelling object detection results which surpass the competitive Fast Faster R-CNN systems. The method described in this paper is the foundation of our submissions to the MS COCO 2015 segmentation competition, where we won the 1st place.",
"We propose a deep convolutional neural network architecture codenamed \"Inception\", which was responsible for setting the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC 2014). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. This was achieved by a carefully crafted design that allows for increasing 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 ILSVRC 2014 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.",
"In this work we train in an end-to-end manner a convolutional neural network (CNN) that jointly handles low-, mid-, and high-level vision tasks in a unified architecture. Such a network can act like a swiss knife for vision tasks, we call it an UberNet to indicate its overarching nature. The main contribution of this work consists in handling challenges that emerge when scaling up to many tasks. We introduce techniques that facilitate (i) training a deep architecture while relying on diverse training sets and (ii) training many (potentially unlimited) tasks with a limited memory budget. This allows us to train in an end-to-end manner a unified CNN architecture that jointly handles (a) boundary detection (b) normal estimation (c) saliency estimation (d) semantic segmentation (e) human part segmentation (f) semantic boundary detection, (g) region proposal generation and object detection. We obtain competitive performance while jointly addressing all tasks in 0.7 seconds on a GPU. Our system will be made publicly available."
]
} |
1704.01074 | 2605133118 | Perception and expression of emotion are key factors to the success of dialogue systems or conversational agents. However, this problem has not been studied in large-scale conversation generation so far. In this paper, we propose Emotional Chatting Machine (ECM) that can generate appropriate responses not only in content (relevant and grammatical) but also in emotion (emotionally consistent). To the best of our knowledge, this is the first work that addresses the emotion factor in large-scale conversation generation. ECM addresses the factor using three new mechanisms that respectively (1) models the high-level abstraction of emotion expressions by embedding emotion categories, (2) captures the change of implicit internal emotion states, and (3) uses explicit emotion expressions with an external emotion vocabulary. Experiments show that the proposed model can generate responses appropriate not only in content but also in emotion. | These works, mainly inspired by psychological findings, are either rule-based, or limited to small data, making them difficult to apply to large-scale conversation generation. Recently, sequence-to-sequence generation models @cite_15 @cite_0 have been successfully applied to large-scale conversation generation @cite_35 , including neural responding machine @cite_23 , hierarchical recurrent models @cite_31 , and many others. These models focus on improving the content quality of the generated responses, including diversity promotion @cite_29 , considering additional information @cite_1 @cite_8 @cite_24 @cite_2 , and handing unknown words @cite_27 . | {
"cite_N": [
"@cite_35",
"@cite_31",
"@cite_8",
"@cite_29",
"@cite_1",
"@cite_0",
"@cite_24",
"@cite_27",
"@cite_23",
"@cite_2",
"@cite_15"
],
"mid": [
"1591706642",
"2296712013",
"2467963359",
"1958706068",
"2521114121",
"2133564696",
"",
"2964165364",
"2159640018",
"2786913767",
"2949888546"
],
"abstract": [
"Conversational modeling is an important task in natural language understanding and machine intelligence. Although previous approaches exist, they are often restricted to specific domains (e.g., booking an airline ticket) and require hand-crafted rules. In this paper, we present a simple approach for this task which uses the recently proposed sequence to sequence framework. Our model converses by predicting the next sentence given the previous sentence or sentences in a conversation. The strength of our model is that it can be trained end-to-end and thus requires much fewer hand-crafted rules. We find that this straightforward model can generate simple conversations given a large conversational training dataset. Our preliminary results suggest that, despite optimizing the wrong objective function, the model is able to converse well. It is able extract knowledge from both a domain specific dataset, and from a large, noisy, and general domain dataset of movie subtitles. On a domain-specific IT helpdesk dataset, the model can find a solution to a technical problem via conversations. On a noisy open-domain movie transcript dataset, the model can perform simple forms of common sense reasoning. As expected, we also find that the lack of consistency is a common failure mode of our model.",
"We consider the task of generative dialogue modeling for movie scripts. To this end, we extend the recently proposed hierarchical recurrent encoder decoder neural network and demonstrate that this model is competitive with state-of-the-art neural language models and backoff n-gram models. We show that its performance can be improved considerably by bootstrapping the learning from a larger questionanswer pair corpus and from pretrained word embeddings.",
"Using neural networks to generate replies in human-computer dialogue systems is attracting increasing attention over the past few years. However, the performance is not satisfactory: the neural network tends to generate safe, universally relevant replies which carry little meaning. In this paper, we propose a content-introducing approach to neural network-based generative dialogue systems. We first use pointwise mutual information (PMI) to predict a noun as a keyword, reflecting the main gist of the reply. We then propose seq2BF, a \"sequence to backward and forward sequences\" model, which generates a reply containing the given keyword. Experimental results show that our approach significantly outperforms traditional sequence-to-sequence models in terms of human evaluation and the entropy measure, and that the predicted keyword can appear at an appropriate position in the reply.",
"Sequence-to-sequence neural network models for generation of conversational responses tend to generate safe, commonplace responses (e.g., \"I don't know\") regardless of the input. We suggest that the traditional objective function, i.e., the likelihood of output (response) given input (message) is unsuited to response generation tasks. Instead we propose using Maximum Mutual Information (MMI) as the objective function in neural models. Experimental results demonstrate that the proposed MMI models produce more diverse, interesting, and appropriate responses, yielding substantive gains in BLEU scores on two conversational datasets and in human evaluations.",
"We consider incorporating topic information into the sequence-to-sequence framework to generate informative and interesting responses for chatbots. To this end, we propose a topic aware sequence-to-sequence (TA-Seq2Seq) model. The model utilizes topics to simulate prior knowledge of human that guides them to form informative and interesting responses in conversation, and leverages the topic information in generation by a joint attention mechanism and a biased generation probability. The joint attention mechanism summarizes the hidden vectors of an input message as context vectors by message attention, synthesizes topic vectors by topic attention from the topic words of the message obtained from a pre-trained LDA model, and let these vectors jointly affect the generation of words in decoding. To increase the possibility of topic words appearing in responses, the model modifies the generation probability of topic words by adding an extra probability item to bias the overall distribution. Empirical study on both automatic evaluation metrics and human annotations shows that TA-Seq2Seq can generate more informative and interesting responses, and significantly outperform the-state-of-the-art response generation models.",
"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.",
"",
"",
"We propose Neural Responding Machine (NRM), a neural network-based response generator for Short-Text Conversation. NRM takes the general encoder-decoder framework: it formalizes the generation of response as a decoding process based on the latent representation of the input text, while both encoding and decoding are realized with recurrent neural networks (RNN). The NRM is trained with a large amount of one-round conversation data collected from a microblogging service. Empirical study shows that NRM can generate grammatically correct and content-wise appropriate responses to over 75 of the input text, outperforming state-of-the-arts in the same setting, including retrieval-based and SMT-based models.",
"",
"Deep Neural Networks (DNNs) are powerful models that have achieved excellent performance on difficult learning tasks. Although DNNs work well whenever large labeled training sets are available, they cannot be used to map sequences to sequences. In this paper, we present a general end-to-end approach to sequence learning that makes minimal assumptions on the sequence structure. Our method uses a multilayered Long Short-Term Memory (LSTM) to map the input sequence to a vector of a fixed dimensionality, and then another deep LSTM to decode the target sequence from the vector. Our main result is that on an English to French translation task from the WMT'14 dataset, the translations produced by the LSTM achieve a BLEU score of 34.8 on the entire test set, where the LSTM's BLEU score was penalized on out-of-vocabulary words. Additionally, the LSTM did not have difficulty on long sentences. For comparison, a phrase-based SMT system achieves a BLEU score of 33.3 on the same dataset. When we used the LSTM to rerank the 1000 hypotheses produced by the aforementioned SMT system, its BLEU score increases to 36.5, which is close to the previous best result on this task. The LSTM also learned sensible phrase and sentence representations that are sensitive to word order and are relatively invariant to the active and the passive voice. Finally, we found that reversing the order of the words in all source sentences (but not target sentences) improved the LSTM's performance markedly, because doing so introduced many short term dependencies between the source and the target sentence which made the optimization problem easier."
]
} |
1704.01074 | 2605133118 | Perception and expression of emotion are key factors to the success of dialogue systems or conversational agents. However, this problem has not been studied in large-scale conversation generation so far. In this paper, we propose Emotional Chatting Machine (ECM) that can generate appropriate responses not only in content (relevant and grammatical) but also in emotion (emotionally consistent). To the best of our knowledge, this is the first work that addresses the emotion factor in large-scale conversation generation. ECM addresses the factor using three new mechanisms that respectively (1) models the high-level abstraction of emotion expressions by embedding emotion categories, (2) captures the change of implicit internal emotion states, and (3) uses explicit emotion expressions with an external emotion vocabulary. Experiments show that the proposed model can generate responses appropriate not only in content but also in emotion. | However, no work has addressed the emotion factor in large-scale conversation generation. There are several studies that generate text from controllable variables. @cite_26 proposed a generative model which can generate sentences conditioned on certain attributes of the language such as sentiment and tenses. Affect Language Model was proposed in @cite_32 to generate text conditioned on context words and affect categories. @cite_22 incorporated the grammar information to generate comments for a document using sentiment and topics. Our work is different in two main aspects: 1) prior studies are heavily dependent on linguistic tools or customized parameters in text generation, while our model is fully data-driven without any manual adjustment; 2) prior studies are unable to model multiple emotion interactions between the input post and the response, instead, the generated text simply continues the emotion of the leading context. | {
"cite_N": [
"@cite_22",
"@cite_26",
"@cite_32"
],
"mid": [
"2739751068",
"2735642330",
"2949378066"
],
"abstract": [
"",
"Generic generation and manipulation of text is challenging and has limited success compared to recent deep generative modeling in visual domain. This paper aims at generating plausible natural language sentences, whose attributes are dynamically controlled by learning disentangled latent representations with designated semantics. We propose a new neural generative model which combines variational auto-encoders and holistic attribute discriminators for effective imposition of semantic structures. With differentiable approximation to discrete text samples, explicit constraints on independent attribute controls, and efficient collaborative learning of generator and discriminators, our model learns highly interpretable representations from even only word annotations, and produces realistic sentences with desired attributes. Quantitative evaluation validates the accuracy of sentence and attribute generation.",
"Human verbal communication includes affective messages which are conveyed through use of emotionally colored words. There has been a lot of research in this direction but the problem of integrating state-of-the-art neural language models with affective information remains an area ripe for exploration. In this paper, we propose an extension to an LSTM (Long Short-Term Memory) language model for generating conversational text, conditioned on affect categories. Our proposed model, Affect-LM enables us to customize the degree of emotional content in generated sentences through an additional design parameter. Perception studies conducted using Amazon Mechanical Turk show that Affect-LM generates naturally looking emotional sentences without sacrificing grammatical correctness. Affect-LM also learns affect-discriminative word representations, and perplexity experiments show that additional affective information in conversational text can improve language model prediction."
]
} |
1704.00763 | 2952326480 | Given a user's query, traditional image search systems rank images according to its relevance to a single modality (e.g., image content or surrounding text). Nowadays, an increasing number of images on the Internet are available with associated meta data in rich modalities (e.g., titles, keywords, tags, etc.), which can be exploited for better similarity measure with queries. In this paper, we leverage visual and textual modalities for image search by learning their correlation with input query. According to the intent of query, attention mechanism can be introduced to adaptively balance the importance of different modalities. We propose a novel Attention guided Multi-modal Correlation (AMC) learning method which consists of a jointly learned hierarchy of intra and inter-attention networks. Conditioned on query's intent, intra-attention networks (i.e., visual intra-attention network and language intra-attention network) attend on informative parts within each modality; a multi-modal inter-attention network promotes the importance of the most query-relevant modalities. In experiments, we evaluate AMC models on the search logs from two real world image search engines and show a significant boost on the ranking of user-clicked images in search results. Additionally, we extend AMC models to caption ranking task on COCO dataset and achieve competitive results compared with recent state-of-the-arts. | Canonical correlation analysis (CCA) @cite_2 learns a cross-modal embedding space to maximize the correlation between different modalities. Kernel CCA (KCCA) @cite_38 extends CCA by adopting a non-linear mapping for different modalities. Alternatively, Nakayama propose kernel principle component analysis with CCA (KPCA-CCA) @cite_4 , which generates input for CCA via non-linear KPCA method. Gong @cite_19 further include a third view into the CCA space by the semantics between image and tags. Similarly, partial least squares (PLS) @cite_39 aims to measure the correlation by projecting multiple sets of data into a latent space. Ngiam @cite_3 introduce deep multimodal learning using neural networks. Recently, Datta @cite_16 first expand to learn the correlation between query and multiple image-related modalities using a graph-based keyphrase extraction model. Compared to @cite_16 , AMC effectively learns a multimodal subspace to measure correlation between query and image-related modalities via three attention networks driven by click-through data. | {
"cite_N": [
"@cite_38",
"@cite_4",
"@cite_3",
"@cite_39",
"@cite_19",
"@cite_2",
"@cite_16"
],
"mid": [
"2156204103",
"",
"2184188583",
"",
"2070753207",
"2100235303",
"2529784951"
],
"abstract": [
"While kernel canonical correlation analysis (CCA) has been applied in many contexts, the convergence of finite sample estimates of the associated functions to their population counterparts has not yet been established. This paper gives a mathematical proof of the statistical convergence of kernel CCA, providing a theoretical justification for the method. The proof uses covariance operators defined on reproducing kernel Hilbert spaces, and analyzes the convergence of their empirical estimates of finite rank to their population counterparts, which can have infinite rank. The result also gives a sufficient condition for convergence on the regularization coefficient involved in kernel CCA: this should decrease as n-1 3, where n is the number of data.",
"",
"Deep networks have been successfully applied to unsupervised feature learning for single modalities (e.g., text, images or audio). In this work, we propose a novel application of deep networks to learn features over multiple modalities. We present a series of tasks for multimodal learning and show how to train deep networks that learn features to address these tasks. In particular, we demonstrate cross modality feature learning, where better features for one modality (e.g., video) can be learned if multiple modalities (e.g., audio and video) are present at feature learning time. Furthermore, we show how to learn a shared representation between modalities and evaluate it on a unique task, where the classifier is trained with audio-only data but tested with video-only data and vice-versa. Our models are validated on the CUAVE and AVLetters datasets on audio-visual speech classification, demonstrating best published visual speech classification on AVLetters and effective shared representation learning.",
"",
"This paper investigates the problem of modeling Internet images and associated text or tags for tasks such as image-to-image search, tag-to-image search, and image-to-tag search (image annotation). We start with canonical correlation analysis (CCA), a popular and successful approach for mapping visual and textual features to the same latent space, and incorporate a third view capturing high-level image semantics, represented either by a single category or multiple non-mutually-exclusive concepts. We present two ways to train the three-view embedding: supervised, with the third view coming from ground-truth labels or search keywords; and unsupervised, with semantic themes automatically obtained by clustering the tags. To ensure high accuracy for retrieval tasks while keeping the learning process scalable, we combine multiple strong visual features and use explicit nonlinear kernel mappings to efficiently approximate kernel CCA. To perform retrieval, we use a specially designed similarity function in the embedded space, which substantially outperforms the Euclidean distance. The resulting system produces compelling qualitative results and outperforms a number of two-view baselines on retrieval tasks on three large-scale Internet image datasets.",
"We present a general method using kernel canonical correlation analysis to learn a semantic representation to web images and their associated text. The semantic space provides a common representation and enables a comparison between the text and images. In the experiments, we look at two approaches of retrieving images based on only their content from a text query. We compare orthogonalization approaches against a standard cross-representation retrieval technique known as the generalized vector space model.",
"Multimodal Retrieval efficiency can be improved by textual query reformulation.A graph based keyphrase extraction incorporating correlation of terms is proposed.Textual query is expanded with relevant part of narratives and extracted keyphrases.Text and image features are combined using a weightlearning model.The proposed method improves both image and text retrieval efficiency significantly. Multimodal Retrieval is a well-established approach for image retrieval. Usually, images are accompanied by text caption along with associated documents describing the image. Textual query expansion as a form of enhancing image retrieval is a relatively less explored area. In this paper, we first study the effect of expanding textual query on both image and its associated text retrieval. Our study reveals that judicious expansion of textual query through keyphrase extraction can lead to better results, either in terms of text-retrieval or both image and text-retrieval. To establish this, we use two well-known keyphrase extraction techniques based on tf-idf and KEA. While query expansion results in increased retrieval efficiency, it is imperative that the expansion be semantically justified. So, we propose a graph-based keyphrase extraction model that captures the relatedness between words in terms of both mutual information and relevance feedback. Most of the existing works have stressed on bridging the semantic gap by using textual and visual features, either in combination or individually. The way these text and image features are combined determines the efficacy of any retrieval. For this purpose, we adopt Fisher-LDA to adjudge the appropriate weights for each modality. This provides us with an intelligent decision-making process favoring the feature set to be infused into the final query. Our proposed algorithm is shown to supersede the previously mentioned keyphrase extraction algorithms for query expansion significantly. A rigorous set of experiments performed on ImageCLEF-2011 Wikipedia Retrieval task dataset validates our claim that capturing the semantic relation between words through Mutual Information followed by expansion of a textual query using relevance feedback can simultaneously enhance both text and image retrieval."
]
} |
1704.00763 | 2952326480 | Given a user's query, traditional image search systems rank images according to its relevance to a single modality (e.g., image content or surrounding text). Nowadays, an increasing number of images on the Internet are available with associated meta data in rich modalities (e.g., titles, keywords, tags, etc.), which can be exploited for better similarity measure with queries. In this paper, we leverage visual and textual modalities for image search by learning their correlation with input query. According to the intent of query, attention mechanism can be introduced to adaptively balance the importance of different modalities. We propose a novel Attention guided Multi-modal Correlation (AMC) learning method which consists of a jointly learned hierarchy of intra and inter-attention networks. Conditioned on query's intent, intra-attention networks (i.e., visual intra-attention network and language intra-attention network) attend on informative parts within each modality; a multi-modal inter-attention network promotes the importance of the most query-relevant modalities. In experiments, we evaluate AMC models on the search logs from two real world image search engines and show a significant boost on the ranking of user-clicked images in search results. Additionally, we extend AMC models to caption ranking task on COCO dataset and achieve competitive results compared with recent state-of-the-arts. | Attention mechanisms have been successfully applied in many computer vision tasks, including object detection @cite_10 and fine-grained image classification @cite_37 . Jin @cite_26 develop an attention-based model for image captioning task that employs an RNN to attend on informative regions in images. Yang @cite_27 and Chen @cite_29 apply attention networks that focus on useful regions in visual question answering (VQA) task. Xiong @cite_36 propose a dynamic memory network to attend on informative visual or textual modality for question answering. Recently, Lu @cite_31 propose a co-attention network to focus on both visual and question modalities in VQA task. Compared to these methods, AMC method not only applies intra-attention networks within each modality, but also employs MTN to balance the importances of modalities based on query's intent for image search task. | {
"cite_N": [
"@cite_37",
"@cite_26",
"@cite_36",
"@cite_29",
"@cite_27",
"@cite_31",
"@cite_10"
],
"mid": [
"",
"1785460851",
"2963579811",
"2174492417",
"2963954913",
"2963668159",
"2951527505"
],
"abstract": [
"",
"Recent progress on automatic generation of image captions has shown that it is possible to describe the most salient information conveyed by images with accurate and meaningful sentences. In this paper, we propose an image caption system that exploits the parallel structures between images and sentences. In our model, the process of generating the next word, given the previously generated ones, is aligned with the visual perception experience where the attention shifting among the visual regions imposes a thread of visual ordering. This alignment characterizes the flow of \"abstract meaning\", encoding what is semantically shared by both the visual scene and the text description. Our system also makes another novel modeling contribution by introducing scene-specific contexts that capture higher-level semantic information encoded in an image. The contexts adapt language models for word generation to specific scene types. We benchmark our system and contrast to published results on several popular datasets. We show that using either region-based attention or scene-specific contexts improves systems without those components. Furthermore, combining these two modeling ingredients attains the state-of-the-art performance.",
"Neural network architectures with memory and attention mechanisms exhibit certain reasoning capabilities required for question answering. One such architecture, the dynamic memory network (DMN), obtained high accuracy on a variety of language tasks. However, it was not shown whether the architecture achieves strong results for question answering when supporting facts are not marked during training or whether it could be applied to other modalities such as images. Based on an analysis of the DMN, we propose several improvements to its memory and input modules. Together with these changes we introduce a novel input module for images in order to be able to answer visual questions. Our new DMN+ model improves the state of the art on both the Visual Question Answering dataset and the bAbI-10k text question-answering dataset without supporting fact supervision.",
"We propose a novel attention based deep learning architecture for visual question answering task (VQA). Given an image and an image related natural language question, VQA generates the natural language answer for the question. Generating the correct answers requires the model's attention to focus on the regions corresponding to the question, because different questions inquire about the attributes of different image regions. We introduce an attention based configurable convolutional neural network (ABC-CNN) to learn such question-guided attention. ABC-CNN determines an attention map for an image-question pair by convolving the image feature map with configurable convolutional kernels derived from the question's semantics. We evaluate the ABC-CNN architecture on three benchmark VQA datasets: Toronto COCO-QA, DAQUAR, and VQA dataset. ABC-CNN model achieves significant improvements over state-of-the-art methods on these datasets. The question-guided attention generated by ABC-CNN is also shown to reflect the regions that are highly relevant to the questions.",
"This paper presents stacked attention networks (SANs) that learn to answer natural language questions from images. SANs use semantic representation of a question as query to search for the regions in an image that are related to the answer. We argue that image question answering (QA) often requires multiple steps of reasoning. Thus, we develop a multiple-layer SAN in which we query an image multiple times to infer the answer progressively. Experiments conducted on four image QA data sets demonstrate that the proposed SANs significantly outperform previous state-of-the-art approaches. The visualization of the attention layers illustrates the progress that the SAN locates the relevant visual clues that lead to the answer of the question layer-by-layer.",
"A number of recent works have proposed attention models for Visual Question Answering (VQA) that generate spatial maps highlighting image regions relevant to answering the question. In this paper, we argue that in addition to modeling \"where to look\" or visual attention, it is equally important to model \"what words to listen to\" or question attention. We present a novel co-attention model for VQA that jointly reasons about image and question attention. In addition, our model reasons about the question (and consequently the image via the co-attention mechanism) in a hierarchical fashion via a novel 1-dimensional convolution neural networks (CNN). Our model improves the state-of-the-art on the VQA dataset from 60.3 to 60.5 , and from 61.6 to 63.3 on the COCO-QA dataset. By using ResNet, the performance is further improved to 62.1 for VQA and 65.4 for COCO-QA.",
"Applying convolutional neural networks to large images is computationally expensive because the amount of computation scales linearly with the number of image pixels. We present a novel recurrent neural network model that is capable of extracting information from an image or video by adaptively selecting a sequence of regions or locations and only processing the selected regions at high resolution. Like convolutional neural networks, the proposed model has a degree of translation invariance built-in, but the amount of computation it performs can be controlled independently of the input image size. While the model is non-differentiable, it can be trained using reinforcement learning methods to learn task-specific policies. We evaluate our model on several image classification tasks, where it significantly outperforms a convolutional neural network baseline on cluttered images, and on a dynamic visual control problem, where it learns to track a simple object without an explicit training signal for doing so."
]
} |
1704.00763 | 2952326480 | Given a user's query, traditional image search systems rank images according to its relevance to a single modality (e.g., image content or surrounding text). Nowadays, an increasing number of images on the Internet are available with associated meta data in rich modalities (e.g., titles, keywords, tags, etc.), which can be exploited for better similarity measure with queries. In this paper, we leverage visual and textual modalities for image search by learning their correlation with input query. According to the intent of query, attention mechanism can be introduced to adaptively balance the importance of different modalities. We propose a novel Attention guided Multi-modal Correlation (AMC) learning method which consists of a jointly learned hierarchy of intra and inter-attention networks. Conditioned on query's intent, intra-attention networks (i.e., visual intra-attention network and language intra-attention network) attend on informative parts within each modality; a multi-modal inter-attention network promotes the importance of the most query-relevant modalities. In experiments, we evaluate AMC models on the search logs from two real world image search engines and show a significant boost on the ranking of user-clicked images in search results. Additionally, we extend AMC models to caption ranking task on COCO dataset and achieve competitive results compared with recent state-of-the-arts. | For image search task, CCA @cite_2 is employed to learn a subspace to maximize correlation between query and image. Ranking CCA (RCCA) @cite_12 refines the CCA space by learning a bilinear ranking function from click-through data. Wang @cite_34 apply a deep ranking model for fine-grained image search and Tan @cite_35 introduce a deep ranking based hashing model. Recently, Gordor @cite_24 apply a region proposal network and Radenovi 'c @cite_7 adopt deep CNN features. Lynch @cite_6 transfer deep semantic features learned from click-through data and apply them on image search task. Compared to the approaches above, AMC method applies VAN to adaptively select informative regions within image modality based on query's intent. On the other side, for textual search task, Joachims @cite_8 introduces click-through data for optimizing search engines. DSSM @cite_30 applies a deep framework to further leverage click-through data. Compared to DSSM @cite_30 , AMC method employs LAN to attend on query-related words. | {
"cite_N": [
"@cite_30",
"@cite_35",
"@cite_7",
"@cite_8",
"@cite_6",
"@cite_24",
"@cite_2",
"@cite_34",
"@cite_12"
],
"mid": [
"2136189984",
"2949235290",
"2336302573",
"2047221353",
"2950339971",
"2340690086",
"2100235303",
"1975517671",
"2216674905"
],
"abstract": [
"Latent semantic models, such as LSA, intend to map a query to its relevant documents at the semantic level where keyword-based matching often fails. In this study we strive to develop a series of new latent semantic models with a deep structure that project queries and documents into a common low-dimensional space where the relevance of a document given a query is readily computed as the distance between them. The proposed deep structured semantic models are discriminatively trained by maximizing the conditional likelihood of the clicked documents given a query using the clickthrough data. To make our models applicable to large-scale Web search applications, we also use a technique called word hashing, which is shown to effectively scale up our semantic models to handle large vocabularies which are common in such tasks. The new models are evaluated on a Web document ranking task using a real-world data set. Results show that our best model significantly outperforms other latent semantic models, which were considered state-of-the-art in the performance prior to the work presented in this paper.",
"With the rapid growth of web images, hashing has received increasing interests in large scale image retrieval. Research efforts have been devoted to learning compact binary codes that preserve semantic similarity based on labels. However, most of these hashing methods are designed to handle simple binary similarity. The complex multilevel semantic structure of images associated with multiple labels have not yet been well explored. Here we propose a deep semantic ranking based method for learning hash functions that preserve multilevel semantic similarity between multi-label images. In our approach, deep convolutional neural network is incorporated into hash functions to jointly learn feature representations and mappings from them to hash codes, which avoids the limitation of semantic representation power of hand-crafted features. Meanwhile, a ranking list that encodes the multilevel similarity information is employed to guide the learning of such deep hash functions. An effective scheme based on surrogate loss is used to solve the intractable optimization problem of nonsmooth and multivariate ranking measures involved in the learning procedure. Experimental results show the superiority of our proposed approach over several state-of-the-art hashing methods in term of ranking evaluation metrics when tested on multi-label image datasets.",
"Convolutional Neural Networks (CNNs) achieve state-of-the-art performance in many computer vision tasks. However, this achievement is preceded by extreme manual annotation in order to perform either training from scratch or fine-tuning for the target task. In this work, we propose to fine-tune CNN for image retrieval from a large collection of unordered images in a fully automated manner. We employ state-of-the-art retrieval and Structure-from-Motion (SfM) methods to obtain 3D models, which are used to guide the selection of the training data for CNN fine-tuning. We show that both hard positive and hard negative examples enhance the final performance in particular object retrieval with compact codes.",
"This paper presents an approach to automatically optimizing the retrieval quality of search engines using clickthrough data. Intuitively, a good information retrieval system should present relevant documents high in the ranking, with less relevant documents following below. While previous approaches to learning retrieval functions from examples exist, they typically require training data generated from relevance judgments by experts. This makes them difficult and expensive to apply. The goal of this paper is to develop a method that utilizes clickthrough data for training, namely the query-log of the search engine in connection with the log of links the users clicked on in the presented ranking. Such clickthrough data is available in abundance and can be recorded at very low cost. Taking a Support Vector Machine (SVM) approach, this paper presents a method for learning retrieval functions. From a theoretical perspective, this method is shown to be well-founded in a risk minimization framework. Furthermore, it is shown to be feasible even for large sets of queries and features. The theoretical results are verified in a controlled experiment. It shows that the method can effectively adapt the retrieval function of a meta-search engine to a particular group of users, outperforming Google in terms of retrieval quality after only a couple of hundred training examples.",
"Search is at the heart of modern e-commerce. As a result, the task of ranking search results automatically (learning to rank) is a multibillion dollar machine learning problem. Traditional models optimize over a few hand-constructed features based on the item's text. In this paper, we introduce a multimodal learning to rank model that combines these traditional features with visual semantic features transferred from a deep convolutional neural network. In a large scale experiment using data from the online marketplace Etsy, we verify that moving to a multimodal representation significantly improves ranking quality. We show how image features can capture fine-grained style information not available in a text-only representation. In addition, we show concrete examples of how image information can successfully disentangle pairs of highly different items that are ranked similarly by a text-only model.",
"We propose a novel approach for instance-level image retrieval. It produces a global and compact fixed-length representation for each image by aggregating many region-wise descriptors. In contrast to previous works employing pre-trained deep networks as a black box to produce features, our method leverages a deep architecture trained for the specific task of image retrieval. Our contribution is twofold: (i) we leverage a ranking framework to learn convolution and projection weights that are used to build the region features; and (ii) we employ a region proposal network to learn which regions should be pooled to form the final global descriptor. We show that using clean training data is key to the success of our approach. To that aim, we use a large scale but noisy landmark dataset and develop an automatic cleaning approach. The proposed architecture produces a global image representation in a single forward pass. Our approach significantly outperforms previous approaches based on global descriptors on standard datasets. It even surpasses most prior works based on costly local descriptor indexing and spatial verification. Additional material is available at www.xrce.xerox.com Deep-Image-Retrieval.",
"We present a general method using kernel canonical correlation analysis to learn a semantic representation to web images and their associated text. The semantic space provides a common representation and enables a comparison between the text and images. In the experiments, we look at two approaches of retrieving images based on only their content from a text query. We compare orthogonalization approaches against a standard cross-representation retrieval technique known as the generalized vector space model.",
"Learning fine-grained image similarity is a challenging task. It needs to capture between-class and within-class image differences. This paper proposes a deep ranking model that employs deep learning techniques to learn similarity metric directly from images. It has higher learning capability than models based on hand-crafted features. A novel multiscale network structure has been developed to describe the images effectively. An efficient triplet sampling algorithm is also proposed to learn the model with distributed asynchronized stochastic gradient. Extensive experiments show that the proposed algorithm outperforms models based on hand-crafted visual features and deep classification models.",
"One of the fundamental problems in image search is to learn the ranking functions, i.e., similarity between the query and image. The research on this topic has evolved through two paradigms: feature-based vector model and image ranker learning. The former relies on the image surrounding texts, while the latter learns a ranker based on human labeled query-image pairs. Each of the paradigms has its own limitation. The vector model is sensitive to the quality of text descriptions, and the learning paradigm is difficult to be scaled up as human labeling is always too expensive to obtain. We demonstrate in this paper that the above two limitations can be well mitigated by jointly exploring subspace learning and the use of click-through data. Specifically, we propose a novel Ranking Canonical Correlation Analysis (RCCA) for learning query and image similarities. RCCA initially finds a common subspace between query and image views by maximizing their correlations, and further simultaneously learns a bilinear query-image similarity function and adjusts the subspace to preserve the preference relations implicit in the click-through data. Once the subspace is finalized, query-image similarity can be computed by the bilinear similarity function on their mappings in this subspace. On a large-scale click-based image dataset with 11.7 million queries and one million images, RCCA is shown to be powerful for image search with superior performance over several state-of-the-art methods on both keyword-based and query-by-example tasks."
]
} |
1704.00485 | 2963867393 | Machine learning (ML) over relational data is a booming area of the database industry and academia. While several projects aim to build scalable and fast ML systems, little work has addressed the pains of sourcing data and features for ML tasks. Real-world relational databases typically have many tables (often, dozens) and data scientists often struggle to even obtain and join all possible tables that provide features for ML. In this context, showed recently that key-foreign key dependencies (KFKDs) between tables often lets us avoid such joins without significantly affecting prediction accuracy--an idea they called avoiding joins safely. While initially controversial, this idea has since been used by multiple companies to reduce the burden of data sourcing for ML. But their work applied only to linear classifiers. In this work, we verify if their results hold for three popular complex classifiers: decision trees, SVMs, and ANNs. We conduct an extensive experimental study using both real-world datasets and simulations to analyze the effects of avoiding KFK joins on such models. Our results show that these high-capacity classifiers are surprisingly and counter-intuitively more robust to avoiding KFK joins compared to linear classifiers, refuting an intuition from the prior work's analysis. We explain this behavior intuitively and identify open questions at the intersection of data management and ML theoretical research. All of our code and datasets are available for download from this http URL. | * Data Integration Integrating data and features from different sources for ML and data mining algorithms often requires applying and adapting techniques from the data integration literature @cite_41 @cite_15 . These include integrating features from different data types in recommendation systems @cite_40 , sensor fusion @cite_2 , dimensionality reduction during feature fusion @cite_29 , and techniques to control data quality during data fusion @cite_43 . Avoiding joins safely can be seen as one schema-based mechanism to reduce the integration burden by predicting a priori if a source table is unlikely to improve ML accuracy. It is a major open challenge to devise similar mechanisms can be devised for other types of data sources, say, using other forms of schema constraints, ontology information, and sampling. There is also a growing interest in making data discovery and other forms of metadata management easier @cite_22 @cite_9 . Our work can be seen as a mechanism to verify the potential utility of some of the discovered data sources using their metadata. We hope our work spurs more research in this direction of exploiting ideas from data integration and data discovery to reduce the data sourcing burden for ML tasks. | {
"cite_N": [
"@cite_22",
"@cite_41",
"@cite_29",
"@cite_9",
"@cite_43",
"@cite_40",
"@cite_2",
"@cite_15"
],
"mid": [
"",
"",
"2012982311",
"",
"2139646386",
"2952846590",
"2038420319",
"1870305865"
],
"abstract": [
"",
"",
"Since the emergence of extensive multimedia data, feature fusion has been more and more important for image and video retrieval, indexing and annotation. Existing feature fusion techniques simply concatenate a pair of different features or use canonical correlation analysis based methods for joint dimensionality reduction in the feature space. However, how to fuse multiple features in a generalized way is still an open problem. In this paper, we reformulate the multiple feature fusion as a general subspace learning problem. The objective of the framework is to find a general linear subspace in which the cumulative pairwise canonical correlation between every pair of feature sets is maximized after the dimension normalization and subspace projection. The learned subspace couples dimensionality reduction and feature fusion together, which can be applied to both unsupervised and supervised learning cases. In the supervised case, the pairwise canonical correlations of feature sets within the same classes are also counted in the objective function for maximization. To better model the high-order feature structure and overcome the computational difficulty, the features extracted from the same pattern source are represented by a single 2D tensor. The tensor-based dimensionality reduction methods are used to further extract low-dimensional discriminative features from the fused feature ensemble. Extensive experiments on visual data classification demonstrate the effectiveness and robustness of the proposed methods.",
"",
"The Big Data era is upon us: data is being generated, collected and analyzed at an unprecedented scale, and data-driven decision making is sweeping through all aspects of society. Since the value of data explodes when it can be linked and fused with other data, addressing the big data integration (BDI) challenge is critical to realizing the promise of Big Data. BDI differs from traditional data integration in many dimensions: (i) the number of data sources, even for a single domain, has grown to be in the tens of thousands, (ii) many of the data sources are very dynamic, as a huge amount of newly collected data are continuously made available, (iii) the data sources are extremely heterogeneous in their structure, with considerable variety even for substantially similar entities, and (iv) the data sources are of widely differing qualities, with significant differences in the coverage, accuracy and timeliness of data provided. This seminar explores the progress that has been made by the data integration community on the topics of schema mapping, record linkage and data fusion in addressing these novel challenges faced by big data integration, and identifies a range of open problems for the community.",
"We demonstrate that, with the availability of distributed computation platforms such as Amazon Web Services and open-source tools, it is possible for a small engineering team to build, launch and maintain a cost-effective, large-scale visual search system with widely available tools. We also demonstrate, through a comprehensive set of live experiments at Pinterest, that content recommendation powered by visual search improve user engagement. By sharing our implementation details and the experiences learned from launching a commercial visual search engines from scratch, we hope visual search are more widely incorporated into today's commercial applications.",
"There has been an ever-increasing interest in multi-disciplinary research on multisensor data fusion technology, driven by its versatility and diverse areas of application. Therefore, there seems to be a real need for an analytical review of recent developments in the data fusion domain. This paper proposes a comprehensive review of the data fusion state of the art, exploring its conceptualizations, benefits, and challenging aspects, as well as existing methodologies. In addition, several future directions of research in the data fusion community are highlighted and described.",
"How do you approach answering queries when your data is stored in multiple databases that were designed independently by different people? This is first comprehensive book on data integration and is written by three of the most respected experts in the field. This book provides an extensive introduction to the theory and concepts underlying today's data integration techniques, with detailed, instruction for their application using concrete examples throughout to explain the concepts. Data integration is the problem of answering queries that span multiple data sources (e.g., databases, web pages). Data integration problems surface in multiple contexts, including enterprise information integration, query processing on the Web, coordination between government agencies and collaboration between scientists. In some cases, data integration is the key bottleneck to making progress in a field. The authors provide a working knowledge of data integration concepts and techniques, giving you the tools you need to develop a complete and concise package of algorithms and applications. *Offers a range of data integration solutions enabling you to focus on what is most relevant to the problem at hand. *Enables you to build your own algorithms and implement your own data integration applications *Companion website with numerous project-based exercises and solutions and slides. Links to commercially available software allowing readers to build their own algorithms and implement their own data integration applications. Facebook page for reader input during and after publication."
]
} |
1704.00412 | 2598160352 | Collaborative crowdsourcing has become a popular approach to organizing work across the globe. Being global also means being vulnerable to shocks -- unforeseen events that disrupt crowds -- that originate from any country. In this study, we examine changes in collaborative behavior of editors of Chinese Wikipedia that arise due to the 2005 government censor- ship in mainland China. Using the exogenous variation in the fraction of editors blocked across different articles due to the censorship, we examine the impact of reduction in group size, which we denote as the shock level, on three collaborative behavior measures: volume of activity, centralization, and conflict. We find that activity and conflict drop on articles that face a shock, whereas centralization increases. The impact of a shock on activity increases with shock level, whereas the impact on centralization and conflict is higher for moderate shock levels than for very small or very high shock levels. These findings provide support for threat rigidity theory -- originally introduced in the organizational theory literature -- in the context of large-scale collaborative crowds. | Threat rigidity has been found to be consistent with behaviors observed in organizations @cite_27 @cite_11 but less so in experimental studies of groups @cite_23 @cite_3 @cite_15 . Other experimental studies found no evidence of centralization under threat @cite_14 @cite_0 . | {
"cite_N": [
"@cite_14",
"@cite_3",
"@cite_0",
"@cite_27",
"@cite_23",
"@cite_15",
"@cite_11"
],
"mid": [
"1978460812",
"",
"2052031610",
"",
"2011854722",
"2163429677",
"2132999934"
],
"abstract": [
"Abstract The effects of uncertainty and threat upon the centralization of group communication structures were investigated in a laboratory study. The performance of these groups was also analyzed. Uncertainty was manipulated by varying stimulus ambiguity; threat was manipulated by varying the possibility of loss and the use of time reminders. Results indicate that group structures became differentiated over time with groups in the high uncertainty, low threat condition showing the greatest degree of centralization. Centralization appeared to be associated with the experience of tension: subjects in the high uncertainty, low threat condition reported the highest degree of tension. Uncertainty had a negative effect on performance. Theoretical and practical implications are discussed.",
"",
"Understanding the processes by which task groups function is a major concern of social psychologists who study small groups. Equally important to the applied group researcher is the understanding of how group processes are altered by the external environment. Research suggests that organizations respond to stress with a centralization of authority so that decision making becomes concentrated in the higher levels of an organizational hierarchy. In this study, this hypothesis was extended to the small-group level of analysis, and the effects of stress on group status and decision making were examined. Experimental results clarify and delimit the centralization-of-authority hypothesis: Rather than centralizing authority and decision making under stress, group leaders and group members become more receptive to information provided by others. Implications for group decision making are discussed.",
"",
"The relationship between external threat and group decision-making processes was tested with 24 groups involved in a management simulation called Tycoon.1 Results indicate a restriction in informat...",
"Teams have become the norm for operating in dangerous and complex situations. To investigate how physical threat affects team performance, 27 threeperson teams engaged in a complex planning and problem-solving task, either under physical threat or under normal conditions. Threat consisted of the possibility that during task performance the oxygen level would be reduced (which, in reality, did not occur). Team processes were measured by automated behavior recordings and questionnaires. Results confirmed that physical threat caused restrictions in information processing, more controlling leadership, less group discussions, and a reduction in coordinating and supporting behavior. These results support the propositions of the threat-rigidity thesis and extend previous research by establishing these results for physical threat and demonstrating effects on coordinating and supporting behavior as well.",
"We develop a theory to explain how individual compassion in response to human pain in organizations becomes socially coordinated through a process we call compassion organizing. The theory specifies five mechanisms, including contextual enabling of attention, emotion, and trust, agents improvising structures, and symbolic enrichment, that show how the social architecture of an organization interacts with agency and emergent features to affect the extraction, generation, coordination, and calibration of resources. In doing so, our theory of compassion organizing suggests that the same structures designed for the normal work of organizations can be redirected to a new purpose to respond to members' pain. We discuss the implications of the theory for compassion organizing and for collective organizing more generally."
]
} |
1704.00515 | 276900368 | Hand motion capture has been an active research topic, following the success of full-body pose tracking. Despite similarities, hand tracking proves to be more challenging, characterized by a higher dimensionality, severe occlusions and self-similarity between fingers. For this reason, most approaches rely on strong assumptions, like hands in isolation or expensive multi-camera systems, that limit practical use. In this work, we propose a framework for hand tracking that can capture the motion of two interacting hands using only a single, inexpensive RGB-D camera. Our approach combines a generative model with collision detection and discriminatively learned salient points. We quantitatively evaluate our approach on 14 new sequences with challenging interactions. | The study of hand motion tracking has its roots in the @math s @cite_20 @cite_46 . Although the problem can be simplified by means of data-gloves @cite_35 , color-gloves @cite_25 , markers @cite_6 or wearable sensors @cite_26 , the ideal solution pursued is the unintrusive, marker-less capture of hand motion. Even until recently the study was mainly confined to the case of a single isolated hand @cite_3 @cite_41 @cite_9 @cite_19 @cite_24 @cite_1 @cite_18 . However, in pursuit of more realistic scenarios, research effort was directed towards the case of a hand interacting with an object @cite_27 @cite_31 @cite_0 , two hands interacting with each other @cite_8 @cite_21 and with an additional object @cite_8 . Multiple objects can be tracked by means of hand tracking and physical forces modeling @cite_23 . | {
"cite_N": [
"@cite_35",
"@cite_41",
"@cite_3",
"@cite_20",
"@cite_18",
"@cite_8",
"@cite_21",
"@cite_23",
"@cite_46",
"@cite_26",
"@cite_6",
"@cite_19",
"@cite_27",
"@cite_25",
"@cite_9",
"@cite_1",
"@cite_24",
"@cite_0",
"@cite_31"
],
"mid": [
"2158662146",
"2138672819",
"2162254475",
"1635989058",
"2148821982",
"2150457612",
"2124419806",
"2130399002",
"2100905386",
"2099800354",
"",
"2100642335",
"2543872873",
"2114663654",
"1533614607",
"2166665167",
"2007104354",
"2153169563",
"1995905186"
],
"abstract": [
"The demand for flexible and re-programmable robots has increased the need for programming by demonstration systems. In this paper, grasp recognition is considered in a programming by demonstration framework. Three methods for grasp recognition are presented and evaluated. The first method uses Hidden Markov Models to model the hand posture sequence during the grasp sequence, while the second method relies on the hand trajectory and hand rotation. The third method is a hybrid method, in which both the first two methods are active in parallel. The particular contribution is that all methods rely on the grasp sequence and not just the final posture of the hand. This facilitates grasp recognition before the grasp is completed. Also, by analyzing the entire sequence and not just the final grasp, the decision is based on more information and increased robustness of the overall system is achieved. The experimental results show that both arm trajectory and final hand posture provide important information for grasp classification. By combining them, the recognition rate of the overall system is increased.",
"In this paper we first describe how we have constructed a 3D deformable Point Distribution Model of the human hand, capturing training data semi-automatically from volume images via a physically-based model. We then show how we have attempted to use this model in tracking an unmarked hand moving with 6 degrees of freedom (plus deformation) in real time using a single video camera. In the course of this we show how to improve on a weighted least-squares pose parameter approximation at little computational cost. We note the successes and shortcomings of our system and discuss how it might be improved.",
"A method is proposed that can generate a ranked list of plausible three-dimensional hand configurations that best match an input image. Hand pose estimation is formulated as an image database indexing problem, where the closest matches for an input hand image are retrieved from a large database of synthetic hand images. In contrast to previous approaches, the system can function in the presence of clutter, thanks to two novel clutter-tolerant indexing methods. First, a computationally efficient approximation of the image-to-model chamfer distance is obtained by embedding binary edge images into a high-dimensional Euclidean space. Second, a general-purpose, probabilistic line matching method identifies those line segment correspondences between model and input images that are the least likely to have occurred by chance. The performance of this clutter tolerant approach is demonstrated in quantitative experiments with hundreds of real hand images.",
"Passive sensing of human hand and limb motion is important for a wide range of applications from human-computer interaction to athletic performance measurement. High degree of freedom articulated mechanisms like the human hand are difficult to track because of their large state space and complex image appearance. This article describes a model-based hand tracking system, called DigitEyes, that can recover the state of a 27 DOF hand model from ordinary gray scale images at speeds of up to 10 Hz.",
"This paper compares two methods for object localization from contours: shape context and chamfer matching of templates. In the light of our experiments, we suggest improvements to the shape context: shape contexts are used to find corresponding features between model and image. In real images it is shown that the shape context is highly influenced by clutters; furthermore, even when the object is correctly localized, the feature correspondence may be poor. We show that the robustness of shape matching can be increased by including a figural continuity constraint. The combined shape and continuity cost is minimized using the Viterbi algorithm on features, resulting in improved localization and correspondence. Our algorithm can be generally applied to any feature based shape matching method. Chamfer matching correlates model templates with the distance transform of the edge image. This can be done efficiently using a coarse-to-fine search over the transformation parameters. The method is robust in clutter, however, multiple templates are needed to handle scale, rotation and shape variation. We compare both methods for locating hand shapes in cluttered images, and applied to word recognition in EZ-Gimpy images.",
"Capturing the motion of two hands interacting with an object is a very challenging task due to the large number of degrees of freedom, self-occlusions, and similarity between the fingers, even in the case of multiple cameras observing the scene. In this paper we propose to use discriminatively learned salient points on the fingers and to estimate the finger-salient point associations simultaneously with the estimation of the hand pose. We introduce a differentiable objective function that also takes edges, optical flow and collisions into account. Our qualitative and quantitative evaluations show that the proposed approach achieves very accurate results for several challenging sequences containing hands and objects in action.",
"We propose a method that relies on markerless visual observations to track the full articulation of two hands that interact with each-other in a complex, unconstrained manner. We formulate this as an optimization problem whose 54-dimensional parameter space represents all possible configurations of two hands, each represented as a kinematic structure with 26 Degrees of Freedom (DoFs). To solve this problem, we employ Particle Swarm Optimization (PSO), an evolutionary, stochastic optimization method with the objective of finding the two-hands configuration that best explains observations provided by an RGB-D sensor. To the best of our knowledge, the proposed method is the first to attempt and achieve the articulated motion tracking of two strongly interacting hands. Extensive quantitative and qualitative experiments with simulated and real world image sequences demonstrate that an accurate and efficient solution of this problem is indeed feasible.",
"In several hand-object(s) interaction scenarios, the change in the objects' state is a direct consequence of the hand's motion. This has a straightforward representation in Newtonian dynamics. We present the first approach that exploits this observation to perform model-based 3D tracking of a table-top scene comprising passive objects and an active hand. Our forward modelling of 3D hand-object(s) interaction regards both the appearance and the physical state of the scene and is parameterized over the hand motion (26 DoFs) between two successive instants in time. We demonstrate that our approach manages to track the 3D pose of all objects and the 3D pose and articulation of the hand by only searching for the parameters of the hand motion. In the proposed framework, covert scene state is inferred by connecting it to the overt state, through the incorporation of physics. Thus, our tracking approach treats a variety of challenging observability issues in a principled manner, without the need to resort to heuristics.",
"Computer sensing of hand and limb motion is an important problem for applications in human computer interaction and computer graphics. We describe a framework for local trading of self occluding motion, in which one part of an object obstructs the visibility of another. Our approach uses a kinematic model to predict occlusions and windowed templates to track partially occluded objects. We present offline 3D tracking results for hand motion with significant self occlusion. >",
"Digits is a wrist-worn sensor that recovers the full 3D pose of the user's hand. This enables a variety of freehand interactions on the move. The system targets mobile settings, and is specifically designed to be low-power and easily reproducible using only off-the-shelf hardware. The electronics are self-contained on the user's wrist, but optically image the entirety of the user's hand. This data is processed using a new pipeline that robustly samples key parts of the hand, such as the tips and lower regions of each finger. These sparse samples are fed into new kinematic models that leverage the biomechanical constraints of the hand to recover the 3D pose of the user's hand. The proposed system works without the need for full instrumentation of the hand (for example using data gloves), additional sensors in the environment, or depth cameras which are currently prohibitive for mobile scenarios due to power and form-factor considerations. We demonstrate the utility of Digits for a variety of application scenarios, including 3D spatial interaction with mobile devices, eyes-free interaction on-the-move, and gaming. We conclude with a quantitative and qualitative evaluation of our system, and discussion of strengths, limitations and future work.",
"",
"We present a novel solution to the problem of recovering and tracking the 3D position, orientation and full articulation of a human hand from markerless visual observations obtained by a Kinect sensor. We treat this as an optimization problem, seeking for the hand model parameters that minimize the discrepancy between the appearance and 3D structure of hypothesized instances of a hand model and actual hand observations. This optimization problem is effectively solved using a variant of Particle Swarm Optimization (PSO). The proposed method does not require special markers and or a complex image acquisition setup. Being model based, it provides continuous solutions to the problem of tracking hand articulations. Extensive experiments with a prototype GPU-based implementation of the proposed method demonstrate that accurate and robust 3D tracking of hand articulations can be achieved in near real-time (15Hz).",
"We present a method for tracking a hand while it is interacting with an object. This setting is arguably the one where hand-tracking has most practical relevance, but poses significant additional challenges: strong occlusions by the object as well as self-occlusions are the norm, and classical anatomical constraints need to be softened due to the external forces between hand and object. To achieve robustness to partial occlusions, we use an individual local tracker for each segment of the articulated structure. The segments are connected in a pairwise Markov random field, which enforces the anatomical hand structure through soft constraints on the joints between adjacent segments. The most likely hand configuration is found with belief propagation. Both range and color data are used as input. Experiments are presented for synthetic data with ground truth and for real data of people manipulating objects.",
"Articulated hand-tracking systems have been widely used in virtual reality but are rarely deployed in consumer applications due to their price and complexity. In this paper, we propose an easy-to-use and inexpensive system that facilitates 3-D articulated user-input using the hands. Our approach uses a single camera to track a hand wearing an ordinary cloth glove that is imprinted with a custom pattern. The pattern is designed to simplify the pose estimation problem, allowing us to employ a nearest-neighbor approach to track hands at interactive rates. We describe several proof-of-concept applications enabled by our system that we hope will provide a foundation for new interactions in modeling, animation control and augmented reality.",
"Partitioned sampling is a technique which was introduced in [I7] for avoiding the high cost of particle filters when tracking more than one object. In fact this technique can reduce the curse of dimensionality in other situations too. This paper describes how to use partitioned sampling on articulated objects, obtaining results that would be impossible with standard sampling methods. Because partitioned sampling is the statistical analogue of a hierarchical search, it makes sense to use it on articulated objects, since links at the base of the object can be localised before moving on to search for subsequent links.",
"This paper presents a practical technique for model-based 3D hand tracking. An anatomically accurate hand model is built from truncated quadrics. This allows for the generation of 2D profiles of the model using elegant tools from projective geometry, and for an efficient method to handle self-occlusion. The pose of the hand model is estimated with an Unscented Kalman filter (UKF), which minimizes the geometric error between the profiles and edges extracted from the images. The use of the UKF permits higher frame rates than more sophisticated estimation methods such as particle filtering, whilst providing higher accuracy than the extended Kalman filter The system is easily scalable from single to multiple views, and from rigid to articulated models. First experiments on real data using one and two cameras demonstrate the quality of the proposed method for tracking a 7 DOF hand model.",
"Tracking the articulated 3D motion of the hand has important applications, for example, in human-computer interaction and teleoperation. We present a novel method that can capture a broad range of articulated hand motions at interactive rates. Our hybrid approach combines, in a voting scheme, a discriminative, part-based pose retrieval method with a generative pose estimation method based on local optimization. Color information from a multi-view RGB camera setup along with a person-specific hand model are used by the generative method to find the pose that best explains the observed images. In parallel, our discriminative pose estimation method uses fingertips detected on depth data to estimate a complete or partial pose of the hand by adopting a part-based pose retrieval strategy. This part-based strategy helps reduce the search space drastically in comparison to a global pose retrieval strategy. Quantitative results show that our method achieves state-of-the-art accuracy on challenging sequences and a near-real time performance of 10 fps on a desktop computer.",
"Due to occlusions, the estimation of the full pose of a human hand interacting with an object is much more challenging than pose recovery of a hand observed in isolation. In this work we formulate an optimization problem whose solution is the 26-DOF hand pose together with the pose and model parameters of the manipulated object. Optimization seeks for the joint hand-object model that (a) best explains the incompleteness of observations resulting from occlusions due to hand-object interaction and (b) is physically plausible in the sense that the hand does not share the same physical space with the object. The proposed method is the first that solves efficiently the continuous, full-DOF, joint hand-object tracking problem based solely on markerless multicamera input. Additionally, it is the first to demonstrate how hand-object interaction can be exploited as a context that facilitates hand pose estimation, instead of being considered as a complicating factor. Extensive quantitative and qualitative experiments with simulated and real world image sequences as well as a comparative evaluation with a state-of-the-art method for pose estimation of isolated hands, support the above findings.",
"In this paper, we propose a prior for hand pose estimation that integrates the direct relation between a manipulating hand and a 3d object. This is of particular interest for a variety of applications since many tasks performed by humans require hand-object interaction. Inspired by the ability of humans to learn the handling of an object from a single example, our focus lies on very sparse training data. We express estimated hand poses in local object coordinates and extract for each individual hand segment, the relative position and orientation as well as contact points on the object. The prior is then modeled as a spatial distribution conditioned to the object. Given a new object of the same object class and new hand dimensions, we can transfer the prior by a procedure involving a geometric warp. In our experiments, we demonstrate that the prior may be used to improve the robustness of a 3d hand tracker and to synthesize a new hand grasping a new object. For this, we integrate the prior into a unified belief propagation framework for tracking and synthesis."
]
} |
1704.00702 | 2592401687 | The design of spacecraft trajectories for missions visiting multiple celestial bodies is here framed as a multi-objective bilevel optimization problem. A comparative study is performed to assess the performance of different Beam Search algorithms at tackling the combinatorial problem of finding the ideal sequence of bodies. Special focus is placed on the development of a new hybridization between Beam Search and the Population-based Ant Colony Optimization algorithm. An experimental evaluation shows all algorithms achieving exceptional performance on a hard benchmark problem. It is found that a properly tuned deterministic Beam Search always outperforms the remaining variants. Beam P-ACO, however, demonstrates lower parameter sensitivity, while offering superior worst-case performance. Being an anytime algorithm, it is then found to be the preferable choice for certain practical applications. | Beam Search @cite_0 @cite_28 has emerged as the standard approach to tackle the combinatorial optimization sub-problems present in most GTOC competitions. Though at times called by other names, it is common to find the general architecture of a tree search that has its computational cost bounded via the selection of a limited number of nodes to branch at each depth-level (non-selected nodes at that depth being immediately discarded). We can find examples of such algorithms in the winning solutions to GTOC4 @cite_23 , GTOC5 @cite_22 , and in the second ranked solution to GTOC7 @cite_6 , which the present research builds on. The Lazy Race Tree Search described in @cite_24 , which at the time presented the best known solution to the GTOC6 problem, can also be seen as a Beam Search variant. In it, the beam'' is composed of all nodes, possibly originating from different tree depths, that fall within a given mission time window. The most promising nodes in that sliding window are branched, and the remaining ones discarded. | {
"cite_N": [
"@cite_22",
"@cite_28",
"@cite_6",
"@cite_0",
"@cite_24",
"@cite_23"
],
"mid": [
"",
"2204575588",
"2225069804",
"",
"2002941233",
"1986159811"
],
"abstract": [
"",
"We discuss the relationships between three approaches to greedy heuristic search: best-first, hill-climbing, and beam search. We consider the design decisions within each family and point out their oft-overlooked similarities. We consider the following best-first searches: weighted A*, greedy search, ASeps, window A* and multi-state commitment k-weighted A*. For hill climbing algorithms, we consider enforced hill climbing and LSS-LRTA*. We also consider a variety of beam searches, including BULB and beam-stack search. We show how to best configure beam search in order to maximize robustness. An empirical analysis on six standard benchmarks reveals that beam search and best-first search have remarkably similar performance, and outperform hill-climbing approaches in terms of both time to solution and solution quality. Of these, beam search is preferable for very large problems and best first search is better on problems where the goal cannot be reached from all states.",
"The design of interplanetary trajectories often involves a preliminary search for options later refined assembled into one final trajectory. It is this broad search that, often being intractable, inspires the international event called Global Trajectory Optimization Competition. In the first part of this chapter, we introduce some fundamental problems of space flight mechanics, building blocks of any attempt to participate successfully in these competitions, and we describe the use of the open source software PyKEP to solve them. In the second part, we formulate an instance of a multiple asteroid rendezvous problem, related to the 7th edition of the competition, and we show step by step how to build a possible solution strategy. In doing so, we introduce two new techniques useful in the design of this particular mission type: the use of an asteroid phasing value and its surrogates and the efficient computation of asteroid clusters. We show how the basic building blocks, sided to these innovative ideas, allow designing an effective global search for possible trajectories.",
"",
"We make use of self-adaptation in a Differential Evolution algorithm and of the asynchronous island model to design a complex interplanetary trajectory touring the Galilean Jupiter moons (Io, Europa, Ganymede and Callisto) using the multiple gravity assist technique. Such a problem was recently the subject of an international competition organized by the Jet Propulsion Laboratory (NASA) and won by a trajectory designed by aerospace experts and reaching the final score of 311 324. We apply our method to the very same problem finding new surprising designs and orbital strategies and a score of up to 316 324.",
"We consider the problem proposed on the 4th Global Trajectory Optimization Competition (GTOC4)."
]
} |
1704.00702 | 2592401687 | The design of spacecraft trajectories for missions visiting multiple celestial bodies is here framed as a multi-objective bilevel optimization problem. A comparative study is performed to assess the performance of different Beam Search algorithms at tackling the combinatorial problem of finding the ideal sequence of bodies. Special focus is placed on the development of a new hybridization between Beam Search and the Population-based Ant Colony Optimization algorithm. An experimental evaluation shows all algorithms achieving exceptional performance on a hard benchmark problem. It is found that a properly tuned deterministic Beam Search always outperforms the remaining variants. Beam P-ACO, however, demonstrates lower parameter sensitivity, while offering superior worst-case performance. Being an anytime algorithm, it is then found to be the preferable choice for certain practical applications. | Evolutionary Algorithms have been explored as an alternative to solve combinatorial problems in mission analysis. In the GTOC5 problem considered here, for instance, @cite_26 and @cite_18 used Genetic Algorithms with hidden genes", to evolve chromosomes encoding asteroid sequences. These approaches were however outperformed in the GTOC5 competition by tree-based approaches. In @cite_25 an evolutionary approach is described for designing debris removal missions. In this highly dynamic trajectory problem, the Inver-over Genetic Algorithm was found to provide competitive solutions to those constructed by different approaches. | {
"cite_N": [
"@cite_18",
"@cite_26",
"@cite_25"
],
"mid": [
"1627997383",
"2291618903",
"2012772810"
],
"abstract": [
"",
"This paper describes the methodologies used to tackle the problem of the 5th Global Trajectory Optimization Competition within the team composed by the Advanced Concepts Team of the European Space Agency and theGlobal Optimization Laboratory of the University of Florence. The method pursued is powered by two innovative approaches: a linearized model of the ‘self fly-by’ aiding a first broad tree search of chemical propulsion options and the use of global optimization techniques (Monotonic Basin Hopping, in this case) applied directly to the low-thrust trajectory model.",
"The space close to our planet is getting more and more polluted. Orbiting debris are posing an increasing threat to operational orbits and the cascading effect, known as Kessler syndrome, may result in a future where the risk of orbiting our planet at some altitudes will be unacceptable. Many argue that the debris density at the Low Earth Orbit (LEO) has already reached a level sufficient to trigger such a cascading effect. An obvious consequence is that we may soon have to actively clean space from debris. Such a space mission will involve a complex combinatorial decision as to choose which debris to remove and in what order. In this paper, we find that this part of the design of an active debris removal mission (ADR) can be mapped into increasingly complex variants to the classic Travelling Salesman Problem (TSP) and that they can be solved by the Inver-over algorithm improving the current state-of-the-art in ADR mission design. We define static and dynamic cases, according to whether we consider the debris orbits as fixed in time or subject to orbital perturbations. We are able, for the first time, to select optimally objects from debris clouds of considerable size: hundreds debris pieces considered while previous works stopped at tens."
]
} |
1704.00702 | 2592401687 | The design of spacecraft trajectories for missions visiting multiple celestial bodies is here framed as a multi-objective bilevel optimization problem. A comparative study is performed to assess the performance of different Beam Search algorithms at tackling the combinatorial problem of finding the ideal sequence of bodies. Special focus is placed on the development of a new hybridization between Beam Search and the Population-based Ant Colony Optimization algorithm. An experimental evaluation shows all algorithms achieving exceptional performance on a hard benchmark problem. It is found that a properly tuned deterministic Beam Search always outperforms the remaining variants. Beam P-ACO, however, demonstrates lower parameter sensitivity, while offering superior worst-case performance. Being an anytime algorithm, it is then found to be the preferable choice for certain practical applications. | A hybridization of Beam Search and ACO was previously presented in @cite_8 . A different hybridization is introduced here, Beam P-ACO'', that differs mainly in the ACO variant under use, and in being a multi-objective algorithm. | {
"cite_N": [
"@cite_8"
],
"mid": [
"2029319864"
],
"abstract": [
"Ant colony optimization (ACO) is a metaheuristic approach to tackle hard combinatorial optimization problems. The basic component of ACO is a probabilistic solution construction mechanism. Due to its constructive nature, ACO can be regarded as a tree search method. Based on this observation, we hybridize the solution construction mechanism of ACO with beam search, which is a well-known tree search method. We call this approach Beam-ACO. The usefulness of Beam-ACO is demonstrated by its application to open shop scheduling (OSS). We experimentally show that Beam-ACO is a state-of-the-art method for OSS by comparing the obtained results to the best available methods on a wide range of benchmark instances."
]
} |
1704.00386 | 2890003917 | Finding the dense regions of a graph and relations among them is a fundamental problem in network analysis. Core and truss decompositions reveal dense subgraphs with hierarchical relations. The incremental nature of algorithms for computing these decompositions and the need for global information at each step of the algorithm hinders scalable parallelization and approximations since the densest regions are not revealed until the end. In a previous work, proposed to iteratively compute the @math -indices of neighbor vertex degrees to obtain the core numbers and prove that the convergence is obtained after a finite number of iterations. This work generalizes the iterative @math -index computation for truss decomposition as well as nucleus decomposition which leverages higher-order structures to generalize core and truss decompositions. In addition, we prove convergence bounds on the number of iterations. We present a framework of local algorithms to obtain the core, truss, and nucleus decompositions. Our algorithms are local, parallel, offer high scalability, and enable approximations to explore time and quality trade-offs. Our shared-memory implementation verifies the efficiency, scalability, and effectiveness of our local algorithms on real-world networks. | Previous attempts to find the approximate core numbers (or @math -cores) focus on the neighborhood of a vertex within a certain radius @cite_46 . It is reported that if the radius is at least half of the diameter, close approximations can be obtained. However, given the small-world nature of the real-world networks, the local graph within a distance of half the diameter is too large to compute. In our work, we approximate the @math -core, @math -truss, and @math nucleus decompositions in a rigorous and efficient way that does not depend on the diameter. | {
"cite_N": [
"@cite_46"
],
"mid": [
"2950201929"
],
"abstract": [
"Graphs are a powerful way to model interactions and relationships in data from a wide variety of application domains. In this setting, entities represented by vertices at the \"center\" of the graph are often more important than those associated with vertices on the \"fringes\". For example, central nodes tend to be more critical in the spread of information or disease and play an important role in clustering community formation. Identifying such \"core\" vertices has recently received additional attention in the context of network experiments , which analyze the response when a random subset of vertices are exposed to a treatment (e.g. inoculation, free product samples, etc). Specifically, the likelihood of having many central vertices in any exposure subset can have a significant impact on the experiment. We focus on using @math -cores and core numbers to measure the extent to which a vertex is central in a graph. Existing algorithms for computing the core number of a vertex require the entire graph as input, an unrealistic scenario in many real world applications. Moreover, in the context of network experiments, the subgraph induced by the treated vertices is only known in a probabilistic sense. We introduce a new method for estimating the core number based only on the properties of the graph within a region of radius @math around the vertex, and prove an asymptotic error bound of our estimator on random graphs. Further, we empirically validate the accuracy of our estimator for small values of @math on a representative corpus of real data sets. Finally, we evaluate the impact of improved local estimation on an open problem in network experimentation posed by"
]
} |
1704.00386 | 2890003917 | Finding the dense regions of a graph and relations among them is a fundamental problem in network analysis. Core and truss decompositions reveal dense subgraphs with hierarchical relations. The incremental nature of algorithms for computing these decompositions and the need for global information at each step of the algorithm hinders scalable parallelization and approximations since the densest regions are not revealed until the end. In a previous work, proposed to iteratively compute the @math -indices of neighbor vertex degrees to obtain the core numbers and prove that the convergence is obtained after a finite number of iterations. This work generalizes the iterative @math -index computation for truss decomposition as well as nucleus decomposition which leverages higher-order structures to generalize core and truss decompositions. In addition, we prove convergence bounds on the number of iterations. We present a framework of local algorithms to obtain the core, truss, and nucleus decompositions. Our algorithms are local, parallel, offer high scalability, and enable approximations to explore time and quality trade-offs. Our shared-memory implementation verifies the efficiency, scalability, and effectiveness of our local algorithms on real-world networks. | Regarding the parallel computations, @cite_15 introduced parallel algorithms to find the number of iterations needed to find the empty @math -core in random hypergraphs. Their work relies on the assumption that the edge density is below a certain threshold and the focus is on the number of iterations only. Our local algorithms present an alternative formulation for the peeling process, and work for any @math value. For the @math -truss decomposition, @cite_48 introduced algorithms for vertex-centric distributed graph processing systems. For the same setup, @cite_43 proposed faster algorithms that can compute @math -trusses in a distributed graph processing system. Both papers make use of the peeling-based algorithms for computation. Our focus is on the local computation where the each edge has access to only its neighbors and no global graph information is necessary, thus promise better scalability. | {
"cite_N": [
"@cite_48",
"@cite_15",
"@cite_43"
],
"mid": [
"2085426754",
"2953245987",
"2070506828"
],
"abstract": [
"Pregel is a system for large scale graph processing developed at Google. It provides a scalable framework for running graph analytics on clusters of commodity machines. In this paper, we present several important undirected graph algorithms for social network analysis which fit within this framework. We discuss various graph componentisation methods, diameter estimation, degrees of separations, along with triangle, k-core and k-truss finding and computing clustering coefficients. Finally we present some experimental results using our own implementation of the Pregel framework, and examine key features of the general framework and algorithmic design.",
"The analysis of several algorithms and data structures can be framed as a peeling process on a random hypergraph: vertices with degree less than k are removed until there are no vertices of degree less than k left. The remaining hypergraph is known as the k-core. In this paper, we analyze parallel peeling processes, where in each round, all vertices of degree less than k are removed. It is known that, below a specific edge density threshold, the k-core is empty with high probability. We show that, with high probability, below this threshold, only (log log n) log(k-1)(r-1) + O(1) rounds of peeling are needed to obtain the empty k-core for r-uniform hypergraphs. Interestingly, we show that above this threshold, Omega(log n) rounds of peeling are required to find the non-empty k-core. Since most algorithms and data structures aim to peel to an empty k-core, this asymmetry appears fortunate. We verify the theoretical results both with simulation and with a parallel implementation using graphics processing units (GPUs). Our implementation provides insights into how to structure parallel peeling algorithms for efficiency in practice.",
"A cohesive subgraph is a primary vehicle for massive graph analysis, and a newly introduced cohesive subgraph, k-truss, which is motivated by a natural observation of social cohesion, has attracted more and more attention. However, the existing parallel solutions to identify the k-truss are inefficient for very large graphs, as they still suffer from huge communication cost and large number of iterations during the computation. In this paper, we propose a novel parallel and efficient truss detection algorithm, called PeTa. The PeTa produces a triangle complete subgraph (TC-subgraph) for every computing node. Based on the TC-subgraphs, PeTa can detect the local k-truss in parallel within a few iterations. We theoretically prove, within this new paradigm, the communication cost of PeTa is bounded by three times of the number of triangles, the total computation complexity of PeTa is the same order as the best known serial algorithm and the number of iterations for a given partition scheme is minimized as well. Furthermore, we present a subgraph-oriented model to efficiently express PeTa in parallel graph computing systems. The results of comprehensive experiments demonstrate, compared with the existing solutions, PeTa saves 2X to 19X in communication cost, reduces 80 to 95 number of iterations and improves the overall performance by 80 across various real-world graphs."
]
} |
1704.00390 | 2949481100 | Deep learning has shown to be effective for robust and real-time monocular image relocalisation. In particular, PoseNet is a deep convolutional neural network which learns to regress the 6-DOF camera pose from a single image. It learns to localize using high level features and is robust to difficult lighting, motion blur and unknown camera intrinsics, where point based SIFT registration fails. However, it was trained using a naive loss function, with hyper-parameters which require expensive tuning. In this paper, we give the problem a more fundamental theoretical treatment. We explore a number of novel loss functions for learning camera pose which are based on geometry and scene reprojection error. Additionally we show how to automatically learn an optimal weighting to simultaneously regress position and orientation. By leveraging geometry, we demonstrate that our technique significantly improves PoseNet's performance across datasets ranging from indoor rooms to a small city. | Large scale localisation research can be divided into two categories; place recognition and metric localisation. Place recognition discretises the world into a number of landmarks and attempts to identify which place is visible in a given image. Traditionally, this has been modelled as an image retrieval problem @cite_51 @cite_55 @cite_4 @cite_53 enabling the use of efficient and scalable retrieval approaches @cite_27 @cite_47 such as Bag-of-Words (BoW) @cite_49 , VLAD @cite_54 @cite_19 , and Fisher vectors @cite_6 . Deep learning models have also been shown to be effective for creating efficient descriptors. Many approaches leverage classification networks @cite_33 @cite_42 @cite_36 @cite_22 , and fine tune them on localisation datasets @cite_14 . Other work of note is PlaNet @cite_45 which trained a classification network to localise images on a world scale. However, all these networks must discretise the world into places and are unable to produce a fine grained estimate of 6-DOF pose. | {
"cite_N": [
"@cite_14",
"@cite_4",
"@cite_33",
"@cite_22",
"@cite_36",
"@cite_55",
"@cite_53",
"@cite_54",
"@cite_42",
"@cite_6",
"@cite_19",
"@cite_27",
"@cite_45",
"@cite_49",
"@cite_47",
"@cite_51"
],
"mid": [
"204268067",
"2013270301",
"2164022341",
"2174726731",
"1833123814",
"2144824356",
"",
"",
"1524680991",
"2012592962",
"2105516263",
"2128017662",
"2284646714",
"2131846894",
"2141362318",
""
],
"abstract": [
"It has been shown that the activations invoked by an image within the top layers of a large convolutional neural network provide a high-level descriptor of the visual content of the image. In this paper, we investigate the use of such descriptors (neural codes) within the image retrieval application. In the experiments with several standard retrieval benchmarks, we establish that neural codes perform competitively even when the convolutional neural network has been trained for an unrelated classification task (e.g. Image-Net). We also evaluate the improvement in the retrieval performance of neural codes, when the network is retrained on a dataset of images that are similar to images encountered at test time.",
"Repeated structures such as building facades, fences or road markings often represent a significant challenge for place recognition. Repeated structures are notoriously hard for establishing correspondences using multi-view geometry. Even more importantly, they violate the feature independence assumed in the bag-of-visual-words representation which often leads to over-counting evidence and significant degradation of retrieval performance. In this work we show that repeated structures are not a nuisance but, when appropriately represented, they form an important distinguishing feature for many places. We describe a representation of repeated structures suitable for scalable retrieval. It is based on robust detection of repeated image structures and a simple modification of weights in the bag-of-visual-word model. Place recognition results are shown on datasets of street-level imagery from Pittsburgh and San Francisco demonstrating significant gains in recognition performance compared to the standard bag-of-visual-words baseline and more recently proposed burstiness weighting.",
"",
"Recently, image representation built upon Convolutional Neural Network (CNN) has been shown to provide effective descriptors for image search, outperforming pre-CNN features as short-vector representations. Yet such models are not compatible with geometry-aware re-ranking methods and still outperformed, on some particular object retrieval benchmarks, by traditional image search systems relying on precise descriptor matching, geometric re-ranking, or query expansion. This work revisits both retrieval stages, namely initial search and re-ranking, by employing the same primitive information derived from the CNN. We build compact feature vectors that encode several image regions without the need to feed multiple inputs to the network. Furthermore, we extend integral images to handle max-pooling on convolutional layer activations, allowing us to efficiently localize matching objects. The resulting bounding box is finally used for image re-ranking. As a result, this paper significantly improves existing CNN-based recognition pipeline: We report for the first time results competing with traditional methods on the challenging Oxford5k and Paris6k datasets.",
"Several recent works have shown that image descriptors produced by deep convolutional neural networks provide state-of-the-art performance for image classification and retrieval problems. It has also been shown that the activations from the convolutional layers can be interpreted as local features describing particular image regions. These local features can be aggregated using aggregation approaches developed for local features (e.g. Fisher vectors), thus providing new powerful global descriptors. In this paper we investigate possible ways to aggregate local deep features to produce compact global descriptors for image retrieval. First, we show that deep features and traditional hand-engineered features have quite different distributions of pairwise similarities, hence existing aggregation methods have to be carefully re-evaluated. Such re-evaluation reveals that in contrast to shallow features, the simple aggregation method based on sum pooling provides arguably the best performance for deep convolutional features. This method is efficient, has few parameters, and bears little risk of overfitting when e.g. learning the PCA matrix. Overall, the new compact global descriptor improves the state-of-the-art on four common benchmarks considerably.",
"This paper describes a probabilistic approach to the problem of recognizing places based on their appearance. The system we present is not limited to localization, but can determine that a new observation comes from a previously unseen place, and so augment its map. Effectively this is a SLAM system in the space of appearance. Our probabilistic approach allows us to explicitly account for perceptual aliasing in the environment—identical but indistinctive observations receive a low probability of having come from the same place. We achieve this by learning a generative model of place appearance. By partitioning the learning problem into two parts, new place models can be learned online from only a single observation of a place. The algorithm complexity is linear in the number of places in the map, and is particularly suitable for online loop closure detection in mobile robotics.",
"",
"",
"Deep convolutional neural networks (CNN) have shown their promise as a universal representation for recognition. However, global CNN activations lack geometric invariance, which limits their robustness for classification and matching of highly variable scenes. To improve the invariance of CNN activations without degrading their discriminative power, this paper presents a simple but effective scheme called multi-scale orderless pooling (MOP-CNN). This scheme extracts CNN activations for local patches at multiple scale levels, performs orderless VLAD pooling of these activations at each level separately, and concatenates the result. The resulting MOP-CNN representation can be used as a generic feature for either supervised or unsupervised recognition tasks, from image classification to instance-level retrieval; it consistently outperforms global CNN activations without requiring any joint training of prediction layers for a particular target dataset. In absolute terms, it achieves state-of-the-art results on the challenging SUN397 and MIT Indoor Scenes classification datasets, and competitive results on ILSVRC2012 2013 classification and INRIA Holidays retrieval datasets.",
"We address the problem of image search on a very large scale, where three constraints have to be considered jointly: the accuracy of the search, its efficiency, and the memory usage of the representation. We first propose a simple yet efficient way of aggregating local image descriptors into a vector of limited dimension, which can be viewed as a simplification of the Fisher kernel representation. We then show how to jointly optimize the dimension reduction and the indexing algorithm, so that it best preserves the quality of vector comparison. The evaluation shows that our approach significantly outperforms the state of the art: the search accuracy is comparable to the bag-of-features approach for an image representation that fits in 20 bytes. Searching a 10 million image dataset takes about 50ms.",
"Recent works on image retrieval have proposed to index images by compact representations encoding powerful local descriptors, such as the closely related VLAD and Fisher vector. By combining such a representation with a suitable coding technique, it is possible to encode an image in a few dozen bytes while achieving excellent retrieval results. This paper revisits some assumptions proposed in this context regarding the handling of \"visual burstiness\", and shows that ad-hoc choices are implicitly done which are not desirable. Focusing on VLAD without loss of generality, we propose to modify several steps of the original design. Albeit simple, these modifications significantly improve VLAD and make it compare favorably against the state of the art.",
"A recognition scheme that scales efficiently to a large number of objects is presented. The efficiency and quality is exhibited in a live demonstration that recognizes CD-covers from a database of 40000 images of popular music CDs. The scheme builds upon popular techniques of indexing descriptors extracted from local regions, and is robust to background clutter and occlusion. The local region descriptors are hierarchically quantized in a vocabulary tree. The vocabulary tree allows a larger and more discriminatory vocabulary to be used efficiently, which we show experimentally leads to a dramatic improvement in retrieval quality. The most significant property of the scheme is that the tree directly defines the quantization. The quantization and the indexing are therefore fully integrated, essentially being one and the same. The recognition quality is evaluated through retrieval on a database with ground truth, showing the power of the vocabulary tree approach, going as high as 1 million images.",
"Is it possible to determine the location of a photo from just its pixels? While the general problem seems exceptionally difficult, photos often contain cues such as landmarks, weather patterns, vegetation, road markings, or architectural details, which in combination allow to infer where the photo was taken. Previously, this problem has been approached using image retrieval methods. In contrast, we pose the problem as one of classification by subdividing the surface of the earth into thousands of multi-scale geographic cells, and train a deep network using millions of geotagged images. We show that the resulting model, called PlaNet, outperforms previous approaches and even attains superhuman accuracy in some cases. Moreover, we extend our model to photo albums by combining it with a long short-term memory (LSTM) architecture. By learning to exploit temporal coherence to geolocate uncertain photos, this model achieves a 50 performance improvement over the single-image model.",
"We describe an approach to object and scene retrieval which searches for and localizes all the occurrences of a user outlined object in a video. The object is represented by a set of viewpoint invariant region descriptors so that recognition can proceed successfully despite changes in viewpoint, illumination and partial occlusion. The temporal continuity of the video within a shot is used to track the regions in order to reject unstable regions and reduce the effects of noise in the descriptors. The analogy with text retrieval is in the implementation where matches on descriptors are pre-computed (using vector quantization), and inverted file systems and document rankings are used. The result is that retrieved is immediate, returning a ranked list of key frames shots in the manner of Google. The method is illustrated for matching in two full length feature films.",
"In this paper, we present a large-scale object retrieval system. The user supplies a query object by selecting a region of a query image, and the system returns a ranked list of images that contain the same object, retrieved from a large corpus. We demonstrate the scalability and performance of our system on a dataset of over 1 million images crawled from the photo-sharing site, Flickr [3], using Oxford landmarks as queries. Building an image-feature vocabulary is a major time and performance bottleneck, due to the size of our dataset. To address this problem we compare different scalable methods for building a vocabulary and introduce a novel quantization method based on randomized trees which we show outperforms the current state-of-the-art on an extensive ground-truth. Our experiments show that the quantization has a major effect on retrieval quality. To further improve query performance, we add an efficient spatial verification stage to re-rank the results returned from our bag-of-words model and show that this consistently improves search quality, though by less of a margin when the visual vocabulary is large. We view this work as a promising step towards much larger, \"web-scale \" image corpora.",
""
]
} |
1704.00390 | 2949481100 | Deep learning has shown to be effective for robust and real-time monocular image relocalisation. In particular, PoseNet is a deep convolutional neural network which learns to regress the 6-DOF camera pose from a single image. It learns to localize using high level features and is robust to difficult lighting, motion blur and unknown camera intrinsics, where point based SIFT registration fails. However, it was trained using a naive loss function, with hyper-parameters which require expensive tuning. In this paper, we give the problem a more fundamental theoretical treatment. We explore a number of novel loss functions for learning camera pose which are based on geometry and scene reprojection error. Additionally we show how to automatically learn an optimal weighting to simultaneously regress position and orientation. By leveraging geometry, we demonstrate that our technique significantly improves PoseNet's performance across datasets ranging from indoor rooms to a small city. | In contrast, metric localisation techniques estimate the metric position and orientation of the camera. Traditionally, this has been approached by computing the pose from 2D-3D correspondences between 2D features in the query image and 3D points in the model, which are determined through descriptor matching @cite_39 @cite_56 @cite_46 @cite_23 @cite_30 . This assumes that the scene is represented by a 3D structure-from-motion model. The full 6 degree-of-freedom pose of a query image can be estimated very precisely @cite_48 . However these methods require a 3D model with a large database of features and efficient retrieval methods. They are expensive to compute, often do not scale well, and are often not robust to changing environmental conditions @cite_28 . | {
"cite_N": [
"@cite_30",
"@cite_28",
"@cite_48",
"@cite_56",
"@cite_39",
"@cite_23",
"@cite_46"
],
"mid": [
"",
"",
"226250005",
"1565312575",
"2102998745",
"153084048",
"1616969904"
],
"abstract": [
"",
"",
"Estimating the absolute pose of a camera relative to a 3D representation of a scene is a fundamental step in many geometric Computer Vision applications. When the camera is calibrated, the pose can be computed very efficiently. If the calibration is unknown, the problem becomes much harder, resulting in slower solvers or solvers requiring more samples and thus significantly longer run-times for RANSAC. In this paper, we challenge the notion that using minimal solvers is always optimal and propose to compute the pose for a camera with unknown focal length by randomly sampling a focal length value and using an efficient pose solver for the now calibrated camera. Our main contribution is a novel sampling scheme that enables us to guide the sampling process towards promising focal length values and avoids considering all possible values once a good pose is found. The resulting RANSAC variant is significantly faster than current state-of-the-art pose solvers, especially for low inlier ratios, while achieving a similar or better pose accuracy.",
"We present a fast, simple location recognition and image localization method that leverages feature correspondence and geometry estimated from large Internet photo collections. Such recovered structure contains a significant amount of useful information about images and image features that is not available when considering images in isolation. For instance, we can predict which views will be the most common, which feature points in a scene are most reliable, and which features in the scene tend to co-occur in the same image. Based on this information, we devise an adaptive, prioritized algorithm for matching a representative set of SIFT features covering a large scene to a query image for efficient localization. Our approach is based on considering features in the scene database, and matching them to query image features, as opposed to more conventional methods that match image features to visual words or database features. We find this approach results in improved performance, due to the richer knowledge of characteristics of the database features compared to query image features. We present experiments on two large city-scale photo collections, showing that our algorithm compares favorably to image retrieval-style approaches to location recognition.",
"Large scale reconstructions of camera matrices and point clouds have been created using structure from motion from community photo collections. Such a dataset is rich in information; it represents a sampling of the geometry and appearance of the underlying space. In this paper, we encode the visibility information between and among points and cameras as visibility probabilities. The conditional visibility probability of a set of points on a point (or a set of cameras on a camera) can rank points (or cameras) based on their mutual dependence. We combine the conditional probability with a distance measure to prioritize points for fast guided search for the image localization problem. We define dual problem of feature triangulation as finding the 3D coordinates of a given image feature point. We use conditional visibility probability to quickly identify a subset of cameras in which a feature is visible.",
"We propose a powerful pipeline for determining the pose of a query image relative to a point cloud reconstruction of a large scene consisting of more than one million 3D points. The key component of our approach is an efficient and effective search method to establish matches between image features and scene points needed for pose estimation. Our main contribution is a framework for actively searching for additional matches, based on both 2D-to-3D and 3D-to-2D search. A unified formulation of search in both directions allows us to exploit the distinct advantages of both strategies, while avoiding their weaknesses. Due to active search, the resulting pipeline is able to close the gap in registration performance observed between efficient search methods and approaches that are allowed to run for multiple seconds, without sacrificing run-time efficiency. Our method achieves the best registration performance published so far on three standard benchmark datasets, with run-times comparable or superior to the fastest state-of-the-art methods.",
"We address the problem of determining where a photo was taken by estimating a full 6-DOF-plus-intrincs camera pose with respect to a large geo-registered 3D point cloud, bringing together research on image localization, landmark recognition, and 3D pose estimation. Our method scales to datasets with hundreds of thousands of images and tens of millions of 3D points through the use of two new techniques: a co-occurrence prior for RANSAC and bidirectional matching of image features with 3D points. We evaluate our method on several large data sets, and show state-of-the-art results on landmark recognition as well as the ability to locate cameras to within meters, requiring only seconds per query."
]
} |
1704.00390 | 2949481100 | Deep learning has shown to be effective for robust and real-time monocular image relocalisation. In particular, PoseNet is a deep convolutional neural network which learns to regress the 6-DOF camera pose from a single image. It learns to localize using high level features and is robust to difficult lighting, motion blur and unknown camera intrinsics, where point based SIFT registration fails. However, it was trained using a naive loss function, with hyper-parameters which require expensive tuning. In this paper, we give the problem a more fundamental theoretical treatment. We explore a number of novel loss functions for learning camera pose which are based on geometry and scene reprojection error. Additionally we show how to automatically learn an optimal weighting to simultaneously regress position and orientation. By leveraging geometry, we demonstrate that our technique significantly improves PoseNet's performance across datasets ranging from indoor rooms to a small city. | In this work, we address the more challenging problem of metric localisation with deep learning. PoseNet @cite_7 introduced the technique of training a convolutional neural network to regress camera pose. It combines the strengths of place recognition and localisation approaches: it can globally relocalise without a good initial pose estimate, and produces a continuous metric pose. Rather than building a map (or database of landmark features), the neural network learns features whose size, unlike a map, does not require memory linearly proportional to the size of the scene. | {
"cite_N": [
"@cite_7"
],
"mid": [
"2951336016"
],
"abstract": [
"We present a robust and real-time monocular six degree of freedom relocalization system. Our system trains a convolutional neural network to regress the 6-DOF camera pose from a single RGB image in an end-to-end manner with no need of additional engineering or graph optimisation. The algorithm can operate indoors and outdoors in real time, taking 5ms per frame to compute. It obtains approximately 2m and 6 degree accuracy for large scale outdoor scenes and 0.5m and 10 degree accuracy indoors. This is achieved using an efficient 23 layer deep convnet, demonstrating that convnets can be used to solve complicated out of image plane regression problems. This was made possible by leveraging transfer learning from large scale classification data. We show the convnet localizes from high level features and is robust to difficult lighting, motion blur and different camera intrinsics where point based SIFT registration fails. Furthermore we show how the pose feature that is produced generalizes to other scenes allowing us to regress pose with only a few dozen training examples. PoseNet code, dataset and an online demonstration is available on our project webpage, at this http URL"
]
} |
1704.00390 | 2949481100 | Deep learning has shown to be effective for robust and real-time monocular image relocalisation. In particular, PoseNet is a deep convolutional neural network which learns to regress the 6-DOF camera pose from a single image. It learns to localize using high level features and is robust to difficult lighting, motion blur and unknown camera intrinsics, where point based SIFT registration fails. However, it was trained using a naive loss function, with hyper-parameters which require expensive tuning. In this paper, we give the problem a more fundamental theoretical treatment. We explore a number of novel loss functions for learning camera pose which are based on geometry and scene reprojection error. Additionally we show how to automatically learn an optimal weighting to simultaneously regress position and orientation. By leveraging geometry, we demonstrate that our technique significantly improves PoseNet's performance across datasets ranging from indoor rooms to a small city. | Later work has extended PoseNet to use RGB-D input @cite_25 , learn relative ego-motion @cite_31 , improve the context of features @cite_28 , localise over video sequences @cite_13 and interpret relocalisation uncertainty with Bayesian Neural Networks @cite_35 . Additionally, @cite_28 demonstrate PoseNet's efficacy on featureless indoor environments, where they demonstrate that SIFT based structure from motion techniques fail in the same environment. | {
"cite_N": [
"@cite_35",
"@cite_28",
"@cite_31",
"@cite_13",
"@cite_25"
],
"mid": [
"2279895976",
"",
"2949948647",
"2952173064",
"2592183487"
],
"abstract": [
"We present a robust and real-time monocular six degree of freedom visual relocalization system. We use a Bayesian convolutional neural network to regress the 6-DOF camera pose from a single RGB image. It is trained in an end-to-end manner with no need of additional engineering or graph optimisation. The algorithm can operate indoors and outdoors in real time, taking under 6ms to compute. It obtains approximately 2m and 6 degrees accuracy for very large scale outdoor scenes and 0.5m and 10 degrees accuracy indoors. Using a Bayesian convolutional neural network implementation we obtain an estimate of the model's relocalization uncertainty and improve state of the art localization accuracy on a large scale outdoor dataset. We leverage the uncertainty measure to estimate metric relocalization error and to detect the presence or absence of the scene in the input image. We show that the model's uncertainty is caused by images being dissimilar to the training dataset in either pose or appearance.",
"",
"This paper presents a convolutional neural network based approach for estimating the relative pose between two cameras. The proposed network takes RGB images from both cameras as input and directly produces the relative rotation and translation as output. The system is trained in an end-to-end manner utilising transfer learning from a large scale classification dataset. The introduced approach is compared with widely used local feature based methods (SURF, ORB) and the results indicate a clear improvement over the baseline. In addition, a variant of the proposed architecture containing a spatial pyramid pooling (SPP) layer is evaluated and shown to further improve the performance.",
"Machine learning techniques, namely convolutional neural networks (CNN) and regression forests, have recently shown great promise in performing 6-DoF localization of monocular images. However, in most cases image-sequences, rather only single images, are readily available. To this extent, none of the proposed learning-based approaches exploit the valuable constraint of temporal smoothness, often leading to situations where the per-frame error is larger than the camera motion. In this paper we propose a recurrent model for performing 6-DoF localization of video-clips. We find that, even by considering only short sequences (20 frames), the pose estimates are smoothed and the localization error can be drastically reduced. Finally, we consider means of obtaining probabilistic pose estimates from our model. We evaluate our method on openly-available real-world autonomous driving and indoor localization datasets.",
"This paper presents an indoor relocalization system using a dual-stream convolutional neural network (CNN) with both color images and depth images as the network inputs. Aiming at the pose regression problem, a deep neural network architecture for RGB-D images is introduced, a training method by stages for the dual-stream CNN is presented, different depth image encoding methods are discussed, and a novel encoding method is proposed. By introducing the range information into the network through a dual-stream architecture, we not only improved the relocalization accuracy by about 20 compared with the state-of-the-art deep learning method for pose regression, but also greatly enhanced the system robustness in challenging scenes such as large-scale, dynamic, fast movement, and night-time environments. To the best of our knowledge, this is the first work to solve the indoor relocalization problems based on deep CNNs with RGB-D camera. The method is first evaluated on the Microsoft 7-Scenes data set to show its advantage in accuracy compared with other CNNs. Large-scale indoor relocalization is further presented using our method. The experimental results show that 0.3 m in position and 4° in orientation accuracy could be obtained. Finally, this method is evaluated on challenging indoor data sets collected from motion capture system. The results show that the relocalization performance is hardly affected by dynamic objects, motion blur, or night-time environments. Note to Practitioners —This paper was motivated by the limitations of the existing indoor relocalization technology that is significant for mobile robot navigation. Using this technology, robots can infer where they are in a previously visited place. Previous visual localization methods can hardly be put into wide application for the reason that they have strict requirements for the environments. When faced with challenging scenes such as large-scale environments, dynamic objects, motion blur caused by fast movement, night-time environments, or other appearance changed scenes, most existing methods tend to fail. This paper introduces deep learning into the indoor relocalization problem and uses dual-stream CNN (depth stream and color stream) to realize 6-DOF pose regression in an end-to-end manner. The localization error is about 0.3 m and 4° in a large-scale indoor environments. And what is more important, the proposed system does not lose efficiency in some challenging scenes. The proposed encoding method of depth images can also be adopted in other deep neural networks with RGB-D cameras as the sensor."
]
} |
1704.00390 | 2949481100 | Deep learning has shown to be effective for robust and real-time monocular image relocalisation. In particular, PoseNet is a deep convolutional neural network which learns to regress the 6-DOF camera pose from a single image. It learns to localize using high level features and is robust to difficult lighting, motion blur and unknown camera intrinsics, where point based SIFT registration fails. However, it was trained using a naive loss function, with hyper-parameters which require expensive tuning. In this paper, we give the problem a more fundamental theoretical treatment. We explore a number of novel loss functions for learning camera pose which are based on geometry and scene reprojection error. Additionally we show how to automatically learn an optimal weighting to simultaneously regress position and orientation. By leveraging geometry, we demonstrate that our technique significantly improves PoseNet's performance across datasets ranging from indoor rooms to a small city. | Although PoseNet is scalable and robust @cite_7 , it does not produce sufficiently accurate estimates of Pose compared to traditional methods @cite_48 . It was designed with a naive regression loss function which trains the network end-to-end without any consideration for geometry. This problem is the focus of this paper -- we do not want to throw away the decades of research into multi view geometry @cite_32 . We improve PoseNet's performance by learning camera pose with a fundamental treatment of scene geometry. | {
"cite_N": [
"@cite_48",
"@cite_32",
"@cite_7"
],
"mid": [
"226250005",
"2033819227",
"2951336016"
],
"abstract": [
"Estimating the absolute pose of a camera relative to a 3D representation of a scene is a fundamental step in many geometric Computer Vision applications. When the camera is calibrated, the pose can be computed very efficiently. If the calibration is unknown, the problem becomes much harder, resulting in slower solvers or solvers requiring more samples and thus significantly longer run-times for RANSAC. In this paper, we challenge the notion that using minimal solvers is always optimal and propose to compute the pose for a camera with unknown focal length by randomly sampling a focal length value and using an efficient pose solver for the now calibrated camera. Our main contribution is a novel sampling scheme that enables us to guide the sampling process towards promising focal length values and avoids considering all possible values once a good pose is found. The resulting RANSAC variant is significantly faster than current state-of-the-art pose solvers, especially for low inlier ratios, while achieving a similar or better pose accuracy.",
"From the Publisher: A basic problem in computer vision is to understand the structure of a real world scene given several images of it. Recent major developments in the theory and practice of scene reconstruction are described in detail in a unified framework. The book covers the geometric principles and how to represent objects algebraically so they can be computed and applied. The authors provide comprehensive background material and explain how to apply the methods and implement the algorithms directly.",
"We present a robust and real-time monocular six degree of freedom relocalization system. Our system trains a convolutional neural network to regress the 6-DOF camera pose from a single RGB image in an end-to-end manner with no need of additional engineering or graph optimisation. The algorithm can operate indoors and outdoors in real time, taking 5ms per frame to compute. It obtains approximately 2m and 6 degree accuracy for large scale outdoor scenes and 0.5m and 10 degree accuracy indoors. This is achieved using an efficient 23 layer deep convnet, demonstrating that convnets can be used to solve complicated out of image plane regression problems. This was made possible by leveraging transfer learning from large scale classification data. We show the convnet localizes from high level features and is robust to difficult lighting, motion blur and different camera intrinsics where point based SIFT registration fails. Furthermore we show how the pose feature that is produced generalizes to other scenes allowing us to regress pose with only a few dozen training examples. PoseNet code, dataset and an online demonstration is available on our project webpage, at this http URL"
]
} |
1704.00693 | 2952668676 | The key common bottleneck in most stencil codes is data movement, and prior research has shown that improving data locality through optimisations that schedule across loops do particularly well. However, in many large PDE applications it is not possible to apply such optimisations through compilers because there are many options, execution paths and data per grid point, many dependent on run-time parameters, and the code is distributed across different compilation units. In this paper, we adapt the data locality improving optimisation called iteration space slicing for use in large OPS applications both in shared-memory and distributed-memory systems, relying on run-time analysis and delayed execution. We evaluate our approach on a number of applications, observing speedups of 2 @math on the Cloverleaf 2D 3D proxy application, which contain 83 141 loops respectively, @math on the linear solver TeaLeaf, and @math on the compressible Navier-Stokes solver OpenSBLI. We demonstrate strong and weak scalability up to 4608 cores of CINECA's Marconi supercomputer. We also evaluate our algorithms on Intel's Knights Landing, demonstrating maintained throughput as the problem size grows beyond 16GB, and we do scaling studies up to 8704 cores. The approach is generally applicable to any stencil DSL that provides per loop data access information. | Tiling by manually modifying code has been demonstrated on smaller codes @cite_44 @cite_35 where one or two loops repeat a large number of times (typically a time iteration); it is a particularly good example of utilising the large caches on CPUs, and they have been studied in detail. | {
"cite_N": [
"@cite_44",
"@cite_35"
],
"mid": [
"2151764765",
"2100580246"
],
"abstract": [
"Stencil computations are an integral component of applications in a number of scientific computing domains. Short-vector SIMD instruction sets are ubiquitous on modern processors and can be used to significantly increase the performance of stencil computations. Traditional approaches to optimizing stencils on these platforms have focused on either short-vector SIMD or data locality optimizations. In this paper, we propose a domain specific language and compiler for stencil computations that allows specification of stencils in a concise manner and automates both locality and short-vector SIMD optimizations, along with effective utilization of multi-core parallelism. Loop transformations to enhance data locality and enable load-balanced parallelism are combined with a data layout transformation to effectively increase the performance of stencil computations. Performance increases are demonstrated for a number of stencils on several modern SIMD architectures.",
"Time skewing and loop tiling has been known for a long time to be a highly beneficial acceleration technique for nested loops especially on bandwidth hungry multi-core processors, but it is little used in practice because efficient implementations utilize complicated code and simple or abstract ones show much smaller gains over naive nested loops. We break this dilemma with an essential time skewing scheme that is both compact and fast."
]
} |
1704.00693 | 2952668676 | The key common bottleneck in most stencil codes is data movement, and prior research has shown that improving data locality through optimisations that schedule across loops do particularly well. However, in many large PDE applications it is not possible to apply such optimisations through compilers because there are many options, execution paths and data per grid point, many dependent on run-time parameters, and the code is distributed across different compilation units. In this paper, we adapt the data locality improving optimisation called iteration space slicing for use in large OPS applications both in shared-memory and distributed-memory systems, relying on run-time analysis and delayed execution. We evaluate our approach on a number of applications, observing speedups of 2 @math on the Cloverleaf 2D 3D proxy application, which contain 83 141 loops respectively, @math on the linear solver TeaLeaf, and @math on the compressible Navier-Stokes solver OpenSBLI. We demonstrate strong and weak scalability up to 4608 cores of CINECA's Marconi supercomputer. We also evaluate our algorithms on Intel's Knights Landing, demonstrating maintained throughput as the problem size grows beyond 16GB, and we do scaling studies up to 8704 cores. The approach is generally applicable to any stencil DSL that provides per loop data access information. | There are a number of compilers focused on applying tiling to stencil computations such as Pochoir @cite_42 , image processing workflows such as Polymage and Halide @cite_7 @cite_0 , and more generally to computations covered by the polyhedral frameworks Pluto @cite_10 @cite_14 , R-STREAM @cite_17 - these have shown significant improvements in performance by exploiting data locality by manipulating loop schedules. There are examples of tiling in distributed memory systems as well: R-STREAM @cite_17 , Pluto @cite_28 , Classen and Griebl @cite_24 . | {
"cite_N": [
"@cite_14",
"@cite_7",
"@cite_28",
"@cite_42",
"@cite_0",
"@cite_24",
"@cite_10",
"@cite_17"
],
"mid": [
"2101285930",
"2084917734",
"",
"1979457157",
"2055312318",
"2103286319",
"2128249697",
""
],
"abstract": [
"Recent advances in polyhedral compilation technology have made it feasible to automatically transform affine sequential loop nests for tiled parallel execution on multi-core processors. However, for multi-statement input programs with statements of different dimensionalities, such as Cholesky or LU decomposition, the parallel tiled code generated by existing automatic parallelization approaches may suffer from significant load imbalance, resulting in poor scalability on multi-core systems. In this paper, we develop a completely automatic parallelization approach for transforming input affine sequential codes into efficient parallel codes that can be executed on a multi-core system in a load-balanced manner. In our approach, we employ a compile-time technique that enables dynamic extraction of inter-tile dependences at run-time, and dynamic scheduling of the parallel tiles on the processor cores for improved scalable execution. Our approach obviates the need for programmer intervention and re-writing of existing algorithms for efficient parallel execution on multi-cores. We demonstrate the usefulness of our approach through comparisons using linear algebra computations: LU and Cholesky decomposition.",
"This paper presents the design and implementation of PolyMage, a domain-specific language and compiler for image processing pipelines. An image processing pipeline can be viewed as a graph of interconnected stages which process images successively. Each stage typically performs one of point-wise, stencil, reduction or data-dependent operations on image pixels. Individual stages in a pipeline typically exhibit abundant data parallelism that can be exploited with relative ease. However, the stages also require high memory bandwidth preventing effective utilization of parallelism available on modern architectures. For applications that demand high performance, the traditional options are to use optimized libraries like OpenCV or to optimize manually. While using libraries precludes optimization across library routines, manual optimization accounting for both parallelism and locality is very tedious. The focus of our system, PolyMage, is on automatically generating high-performance implementations of image processing pipelines expressed in a high-level declarative language. Our optimization approach primarily relies on the transformation and code generation capabilities of the polyhedral compiler framework. To the best of our knowledge, this is the first model-driven compiler for image processing pipelines that performs complex fusion, tiling, and storage optimization automatically. Experimental results on a modern multicore system show that the performance achieved by our automatic approach is up to 1.81x better than that achieved through manual tuning in Halide, a state-of-the-art language and compiler for image processing pipelines. For a camera raw image processing pipeline, our performance is comparable to that of a hand-tuned implementation.",
"",
"A stencil computation repeatedly updates each point of a d-dimensional grid as a function of itself and its near neighbors. Parallel cache-efficient stencil algorithms based on \"trapezoidal decompositions\" are known, but most programmers find them difficult to write. The Pochoir stencil compiler allows a programmer to write a simple specification of a stencil in a domain-specific stencil language embedded in C++ which the Pochoir compiler then translates into high-performing Cilk code that employs an efficient parallel cache-oblivious algorithm. Pochoir supports general d-dimensional stencils and handles both periodic and aperiodic boundary conditions in one unified algorithm. The Pochoir system provides a C++ template library that allows the user's stencil specification to be executed directly in C++ without the Pochoir compiler (albeit more slowly), which simplifies user debugging and greatly simplified the implementation of the Pochoir compiler itself. A host of stencil benchmarks run on a modern multicore machine demonstrates that Pochoir outperforms standard parallelloop implementations, typically running 2-10 times faster. The algorithm behind Pochoir improves on prior cache-efficient algorithms on multidimensional grids by making \"hyperspace\" cuts, which yield asymptotically more parallelism for the same cache efficiency.",
"Image processing pipelines combine the challenges of stencil computations and stream programs. They are composed of large graphs of different stencil stages, as well as complex reductions, and stages with global or data-dependent access patterns. Because of their complex structure, the performance difference between a naive implementation of a pipeline and an optimized one is often an order of magnitude. Efficient implementations require optimization of both parallelism and locality, but due to the nature of stencils, there is a fundamental tension between parallelism, locality, and introducing redundant recomputation of shared values. We present a systematic model of the tradeoff space fundamental to stencil pipelines, a schedule representation which describes concrete points in this space for each stage in an image processing pipeline, and an optimizing compiler for the Halide image processing language that synthesizes high performance implementations from a Halide algorithm and a schedule. Combining this compiler with stochastic search over the space of schedules enables terse, composable programs to achieve state-of-the-art performance on a wide range of real image processing pipelines, and across different hardware architectures, including multicores with SIMD, and heterogeneous CPU+GPU execution. From simple Halide programs written in a few hours, we demonstrate performance up to 5x faster than hand-tuned C, intrinsics, and CUDA implementations optimized by experts over weeks or months, for image processing applications beyond the reach of past automatic compilers.",
"The polytope model has been used successfully as a tool for program analysis and transformation in the field of automatic loop parallelization. However, for the final step of automatic code generation, the generated code is either only usable on shared memory architectures or severely restricts the parallelization methods that can be applied. In this paper, we present a fully automated method for generating efficient target code, which is executable on clusters that are based on a distributed memory architecture. We also provide speedup results of experiments on a local cluster.",
"The polyhedral model provides powerful abstractions to optimize loop nests with regular accesses. Affine transformations in this model capture a complex sequence of execution-reordering loop transformations that can improve performance by parallelization as well as locality enhancement. Although a significant body of research has addressed affine scheduling and partitioning, the problem of automaticallyfinding good affine transforms forcommunication-optimized coarsegrained parallelization together with locality optimization for the general case of arbitrarily-nested loop sequences remains a challenging problem. We propose an automatic transformation framework to optimize arbitrarilynested loop sequences with affine dependences for parallelism and locality simultaneously. The approach finds good tiling hyperplanes by embedding a powerful and versatile cost function into an Integer Linear Programming formulation. These tiling hyperplanes are used for communication-minimized coarse-grained parallelization as well as for locality optimization. The approach enables the minimization of inter-tile communication volume in the processor space, and minimization of reuse distances for local execution at each node. Programs requiring one-dimensional versusmulti-dimensional time schedules (with scheduling-based approaches) are all handled with the same algorithm. Synchronization-free parallelism, permutable loops or pipelined parallelismat various levels can be detected. Preliminary studies of the framework show promising results.",
""
]
} |
1704.00693 | 2952668676 | The key common bottleneck in most stencil codes is data movement, and prior research has shown that improving data locality through optimisations that schedule across loops do particularly well. However, in many large PDE applications it is not possible to apply such optimisations through compilers because there are many options, execution paths and data per grid point, many dependent on run-time parameters, and the code is distributed across different compilation units. In this paper, we adapt the data locality improving optimisation called iteration space slicing for use in large OPS applications both in shared-memory and distributed-memory systems, relying on run-time analysis and delayed execution. We evaluate our approach on a number of applications, observing speedups of 2 @math on the Cloverleaf 2D 3D proxy application, which contain 83 141 loops respectively, @math on the linear solver TeaLeaf, and @math on the compressible Navier-Stokes solver OpenSBLI. We demonstrate strong and weak scalability up to 4608 cores of CINECA's Marconi supercomputer. We also evaluate our algorithms on Intel's Knights Landing, demonstrating maintained throughput as the problem size grows beyond 16GB, and we do scaling studies up to 8704 cores. The approach is generally applicable to any stencil DSL that provides per loop data access information. | The kinds of transformations applied are also wide-ranging, starting at the simplest skewed tiling methods across time iterations @cite_35 @cite_23 , wavefront methods @cite_23 @cite_21 , and their combinations with various tile shapes such as diamond and hexagonal tiling @cite_18 @cite_2 . | {
"cite_N": [
"@cite_35",
"@cite_18",
"@cite_21",
"@cite_23",
"@cite_2"
],
"mid": [
"2100580246",
"1506424797",
"2160106616",
"2038481157",
"2166622045"
],
"abstract": [
"Time skewing and loop tiling has been known for a long time to be a highly beneficial acceleration technique for nested loops especially on bandwidth hungry multi-core processors, but it is little used in practice because efficient implementations utilize complicated code and simple or abstract ones show much smaller gains over naive nested loops. We break this dilemma with an essential time skewing scheme that is both compact and fast.",
"The importance of stencil-based algorithms in computational science has focused attention on optimized parallel implementations for multilevel cache-based processors. Temporal blocking schemes leverage the large bandwidth and low latency of caches to accelerate stencil updates and approach theoretical peak performance. A key ingredient is the reduction of data traffic across slow data paths, especially the main memory interface. In this work we combine the ideas of multicore wavefront temporal blocking and diamond tiling to arrive at stencil update schemes that show large reductions in memory pressure compared to existing approaches. The resulting schemes show performance advantages in bandwidth-starved situations, which are exacerbated by the high bytes per lattice update case of variable coefficients. Our thread groups concept provides a controllable trade-off between concurrency and memory usage, shifting the pressure between the memory interface and the CPU. We present performance results on a contemp...",
"Time skewing is a compile-time optimization that can provide arbitrarily high cache hit rates for a class of iterative calculations, given a sufficient number of time steps and sufficient cache memory. Thus, it can eliminate processor idle time caused by inadequate main memory bandwidth. In this article, we give a generalization of time skewing for multiprocessor architectures, and discuss time skewing for multilevel caches. Our generalization for multiprocessors lets us eliminate processor idle time caused by any combination of inadequate main memory bandwidth, limited network bandwidth, and high network latency, given a sufficiently large problem and sufficient cache. As in the uniprocessor case, the cache requirement grows with the machine balance rather than the problem size. Our techniques for using multilevel caches reduce the LI cache requirement, which would otherwise be unacceptably high for some architectures when using arrays of high dimension.",
"Loop skewing is a new procedure to derive the wavefront method of execution of nested loops. The wavefront method is used to execute nested loops on parallel and vector computers when none of the loops can be done in vector mode. Loop skewing is a simple transformation of loop bounds and is combined with loop interchanging to generate the wavefront. This derivation is particularly suitable for implementation in compilers that already perform automatic detection of parallelism and generation of vector and parallel code, such as are available today. Loop normalization, a loop transformation used by several vectorizing translators, is related to loop skewing, and we show how loop normalization, applied blindly, can adversely affect the parallelism detected by these translators.",
"Iterative stencil computations are important in scientic computing and more and more also in the embedded and mobile domain. Recent publications have shown that tiling schemes that ensure concurrent start provide ecient ways to execute these kernels. Diamond tiling and hybrid-hexagonal tiling are two successful tiling schemes that enable concurrent start. Both have dierent advantages: diamond tiling is integrated in a general purpose optimization framework and uses a cost function to choose among tiling hyperplanes, whereas the more exible tile sizes of hybrid-hexagonal tiling have proven to be eective for the generation of GPU code. We show that these two approaches are even more interesting when combined. We revisit the formalization of diamond and hexagonal tiling, present the eects of tile size and wavefront choices on tile-level parallelism, and formulate constraints for optimal diamond tile shapes. We then extend the diamond tiling formulation into a hexagonal tiling one, combining the benets of both. The paper closes with an outlook of hexagonal tiling in higher dimensional spaces, an important generalization suitable for massively parallel architectures."
]
} |
1704.00693 | 2952668676 | The key common bottleneck in most stencil codes is data movement, and prior research has shown that improving data locality through optimisations that schedule across loops do particularly well. However, in many large PDE applications it is not possible to apply such optimisations through compilers because there are many options, execution paths and data per grid point, many dependent on run-time parameters, and the code is distributed across different compilation units. In this paper, we adapt the data locality improving optimisation called iteration space slicing for use in large OPS applications both in shared-memory and distributed-memory systems, relying on run-time analysis and delayed execution. We evaluate our approach on a number of applications, observing speedups of 2 @math on the Cloverleaf 2D 3D proxy application, which contain 83 141 loops respectively, @math on the linear solver TeaLeaf, and @math on the compressible Navier-Stokes solver OpenSBLI. We demonstrate strong and weak scalability up to 4608 cores of CINECA's Marconi supercomputer. We also evaluate our algorithms on Intel's Knights Landing, demonstrating maintained throughput as the problem size grows beyond 16GB, and we do scaling studies up to 8704 cores. The approach is generally applicable to any stencil DSL that provides per loop data access information. | The only works we are aware of that has applied similar transformations to large-scale scientific problems are the Formura DSL @cite_13 , which is in full control of the code that is being generated from high-level mathematical expressions - therefore it avoids the issue of various execution paths and multiple compilation units to tile across. Work by Malas et. al @cite_12 applied a combination of wavefront and diamond tiling to an electromagnetics code, however only a handful of loops are tiled across, and it is done mostly by hand. | {
"cite_N": [
"@cite_13",
"@cite_12"
],
"mid": [
"2562159580",
"2196501564"
],
"abstract": [
"Stencil computation has many applications in science and engineering, thus many optimization techniques such as temporal blocking have been developed. They are, however, rarely used in real-world applications, since a large amount of careful programming is required for even the simplest of stencils. We introduce Formura, a domain specific language that provides easy access to optimized stencil computations. Higher-order integration schemes can be defined using mathematical notations. Formura generates C code with MPI calls and performs autotuning. Hence its performance is portable to most distributed-memory computers. We show the scientific applicability of Formura by performing magnetohydrodynamics (MHD) and below-ground biology simulations. Ability to reach bytes-per-flops ratio only attainable by temporal blocking is demonstrated. We also demonstrate scaling up to the full nodes of the K computer, with 1.184 Pflops, 11.62 floating-point-operation efficiency, and 31.26 memory throughput efficiency.",
"Understanding and optimizing the properties of solar cells is becoming a key issue in the search for alternatives to nuclear and fossil energy sources. A theoretical analysis via numerical simulations involves solving Maxwell's Equations in discretized form and typically requires substantial computing effort. We start from a hybrid-parallel (MPI+OpenMP) production code that implements the Time Harmonic Inverse Iteration Method (THIIM) with Finite-Difference Frequency Domain (FDFD) discretization. Although this algorithm has the characteristics of a strongly bandwidth-bound stencil update scheme, it is significantly different from the popular stencil types that have been exhaustively studied in the high performance computing literature to date. We apply a recently developed stencil optimization technique, multicore wavefront diamond tiling with multi-dimensional cache block sharing, and describe in detail the peculiarities that need to be considered due to the special stencil structure. Concurrency in updating the components of the electric and magnetic fields provides an additional level of parallelism. The dependence of the cache size requirement of the optimized code on the blocking parameters is modeled accurately, and an auto-tuner searches for optimal configurations in the remaining parameter space. We were able to completely decouple the execution from the memory bandwidth bottleneck, accelerating the implementation by a factor of three to four compared to an optimal implementation with pure spatial blocking on an 18-core Intel Haswell CPU."
]
} |
1704.00693 | 2952668676 | The key common bottleneck in most stencil codes is data movement, and prior research has shown that improving data locality through optimisations that schedule across loops do particularly well. However, in many large PDE applications it is not possible to apply such optimisations through compilers because there are many options, execution paths and data per grid point, many dependent on run-time parameters, and the code is distributed across different compilation units. In this paper, we adapt the data locality improving optimisation called iteration space slicing for use in large OPS applications both in shared-memory and distributed-memory systems, relying on run-time analysis and delayed execution. We evaluate our approach on a number of applications, observing speedups of 2 @math on the Cloverleaf 2D 3D proxy application, which contain 83 141 loops respectively, @math on the linear solver TeaLeaf, and @math on the compressible Navier-Stokes solver OpenSBLI. We demonstrate strong and weak scalability up to 4608 cores of CINECA's Marconi supercomputer. We also evaluate our algorithms on Intel's Knights Landing, demonstrating maintained throughput as the problem size grows beyond 16GB, and we do scaling studies up to 8704 cores. The approach is generally applicable to any stencil DSL that provides per loop data access information. | Identifying the sequence of loops to tile across and to carry out dependency analysis is a lot easier at run-time, particularly with the help of delayed evaluation or lazy execution @cite_38 @cite_11 , which is a well-known technique used particularly in functional languages that allows expressions to be evaluated only when their results are required. Lazy execution is also used in other fields, such as Apache Spark to plan out the sequence of computations and to skip unnecessary steps. We apply the lazy execution idea to figure out dependencies and compute loops schedules at runtime - to our knowledge these two have not been used together in scientific computing. | {
"cite_N": [
"@cite_38",
"@cite_11"
],
"mid": [
"2067911767",
"1984646543"
],
"abstract": [
"A different way to execute pure LISP programs is presented. It delays the evaluation of parameters and list structures without ever having to perform more evaluation steps than the usual method. Although the central idea can be found in earlier work this paper is of interest since it treats a rather well-known language and works out an algorithm which avoids full substitution. A partial correctness proof using Scott-Strachey semantics is sketched in a later section.",
"Implementations of lazy evaluation for nonstrict functional languages usually involve the notion of a delayed representation of the value of an expression, which we call athunk. We present several techniques for implementing thunks and formalize a class of optimizations that reduce both the space and time overhead of these techniques. The optimizations depend on a compile-time inferencing strategy calledpath analysis, a generalization of strictness analysis that uncovers order-of-evaluation information. Although the techniques in this paper are focused on the compilation of a nonstrict functional language for a conventional architecture, they are directly applicable to most of the virtual machines commonly used for implementing such languages. The same techniques also apply to other forms of delayed evaluation such asfutures andpromises."
]
} |
1704.00497 | 2950536460 | UPGMA (Unweighted Pair Group Method with Arithmetic Mean) is a widely used clustering method. Here we show that UPGMA is a greedy heuristic for the normalized equidistant minimum evolution (NEME) problem, that is, finding a rooted tree that minimizes the minimum evolution score relative to the dissimilarity matrix among all rooted trees with the same leaf-set in which all leaves have the same distance to the root. We prove that the NEME problem is NP-hard. In addition, we present some heuristic and approximation algorithms for solving the NEME problem, including a polynomial time algorithm that yields a binary, rooted tree whose NEME score is within O(log^2 n) of the optimum. We expect that these results to eventually provide further insights into the behavior of the UPGMA algorithm. | Another recent, related direction of work considers the algebraic structure of the space of rooted phylogenetic trees induced by the UPGMA method (see, e.g. @cite_17 @cite_16 ). This algebraic structure is tightly linked with the property of consistency of a tree construction method, that is, those conditions under which the method is able to reconstruct a tree that has been used to generate the input dissimilarity matrix (see, e.g. @cite_12 ). In the context of our work, we are particularly interested in the consistency of methods that perform a local search of the space of all rooted phylogenetic trees on a fixed set of leaves (see, e.g. @cite_19 ). Again, balanced minimum evolution is the variant of minimum evolution for which some consistency results of this type are known @cite_25 @cite_23 . | {
"cite_N": [
"@cite_19",
"@cite_23",
"@cite_16",
"@cite_25",
"@cite_12",
"@cite_17"
],
"mid": [
"2029423003",
"1986154345",
"1847059380",
"2129696562",
"2042263183",
"2047401539"
],
"abstract": [
"Due to its speed, the distance approach remains the best hope for building phylogenies on very large sets of taxa. Recently (R. Desper and O. Gascuel, J. Comp. Biol. 9:687-705, 2002), we introduced a new \"balanced\" minimum evolution (BME) principle, based on a branch length estimation scheme of Y. Pauplin (J. Mol. Evol. 51:41-47, 2000). Initial simulations suggested that FASTME, our program implementing the BME principle, was more accurate than or equivalent to all other distance methods we tested, with running time significantly faster than Neighbor-Joining (NJ). This article further explores the properties of the BME principle, and it explains and illustrates its impressive topological accuracy. We prove that the BME principle is a special case of the weighted least-squares approach, with biologically meaningful variances of the distance estimates. We show that the BME principle is statistically consistent. We demonstrate that FASTME only produces trees with positive branch lengths, a feature that separates this approach from NJ (and related methods) that may produce trees with branches with biologically meaningless negative lengths. Finally, we consider a large simulated data set, with 5,000 100-taxon trees generated by the Aldous beta-splitting distribution encompassing a range of distributions from Yule-Harding to uniform, and using a covarion-like model of sequence evolution. FASTME produces trees faster than NJ, and much faster than WEIGHBOR and the weighted least-squares implementation of PAUP*. Moreover, FASTME trees are consistently more accurate at all settings, ranging from Yule-Harding to uniform distributions, and all ranges of maximum pairwise divergence and departure from molecular clock. Interestingly, the covarion parameter has little effect on the tree quality for any of the algorithms. FASTME is freely available on the web.",
"Distance-based phylogenetic methods attempt to reconstruct an accurate phylogenetic tree from an estimated matrix of pairwise distances between taxa. This paper examines two distance-based algorithms (GreedyBME and FastME) that are based on the principle of minimizing the balanced minimum evolution score of the output tree in relation to the given estimated distance matrix. This is also the principle that underlies the neighbor-joining (NJ) algorithm. We show that GreedyBME and FastME both reconstruct the entire correct tree if the input data are quartet consistent, and also that if the maximum error of any distance estimate is @math , then both algorithms output trees containing all sufficiently long edges of the true tree: those having length at least @math . That is to say, the algorithms have edge safety radius 1 3. In contrast, quartet consistency of the data is not sufficient to guarantee the NJ algorithm reconstructs the correct tree, and moreover, the NJ algorithm has edge safety radius of 1 4: Only edges of the true tree of length at least @math can be guaranteed to appear in the output. These results give further theoretical support to the experimental evidence suggesting FastME is a more suitable distance-based phylogeny reconstruction method than the NJ algorithm.",
"UPGMA is a heuristic method identifying the least squares equidistant phylogenetic tree given empirical distance data among @math taxa. We study this classic algorithm using the geometry of the space of all equidistant trees with @math leaves, also known as the Bergman complex of the graphical matroid for the complete graph @math . We show that UPGMA performs an orthogonal projection of the data onto a maximal cell of the Bergman complex. We also show that the equidistant tree with the least (Euclidean) distance from the data is obtained from such an orthogonal projection, but not necessarily given by UPGMA. Using this geometric information we give an extension of the UPGMA algorithm. We also present a branch and bound method for finding the best equidistant tree. Finally, we prove that there are distance data among @math taxa which project to at least @math equidistant trees.",
"Many phylogenetic algorithms search the space of possible trees using topological rearrangements and some optimality criterion. FastME is such an approach that uses the balanced minimum evolution (BME) principle, which computer studies have demonstrated to have high accuracy. FastME includes two variants: balanced subtree prune and regraft (BSPR) and balanced nearest neighbor interchange (BNNI) . These algorithms take as input a distance matrix and a putative phylogenetic tree. The tree is modified using SPR or NNI operations, respectively, to reduce the BME length relative to the distance matrix, until a tree with (locally) shortest BME length is found. Following computer simulations, it has been conjectured that BSPR and BNNI are consistent, i.e. for an input distance that is a tree-metric, they converge to the corresponding tree. We prove that the BSPR algorithm is consistent. Moreover, even if the input contains small errors relative to a tree-metric, we show that the BSPR algorithm still returns the corresponding tree. Whether BNNI is consistent remains open.",
"Several popular methods for phylogenetic inference (or hierarchical clustering) are based on a matrix of pairwise distances between taxa (or any kind of objects): The objective is to construct a tree with branch lengths so that the distances between the leaves in that tree are as close as possible to the input distances. If we hold the structure (topology) of the tree fixed, in some relevant cases (e.g., ordinary least squares) the optimal values for the branch lengths can be expressed using simple combinatorial formulae. Here we define a general form for these formulae and show that they all have two desirable properties: First, the common tree reconstruction approaches (least squares, minimum evolution), when used in combination with these formulae, are guaranteed to infer the correct tree when given enough data (consistency); second, the branch lengths of all the simple (nearest neighbor interchange) rearrangements of a tree can be calculated, optimally, in quadratic time in the size of the tree, thus allowing the efficient application of hill climbing heuristics. The study presented here is a continuation of that by Mihaescu and Pachter on branch length estimation [Mihaescu R, Pachter L (2008) Proc Natl Acad Sci USA 105:13206–13211]. The focus here is on the inference of the tree itself and on providing a basis for novel algorithms to reconstruct trees from distances.",
"Distance-based methods such as UPGMA (Unweighted Pair Group Method with Arithmetic Mean) continue to play a significant role in phylogenetic research. We use polyhedral combinatorics to analyze the natural subdivision of the positive orthant induced by classifying the input vectors according to tree topologies returned by the algorithm. The partition lattice informs the study of UPGMA trees. We give a closed form for the extreme rays of UPGMA cones on n taxa, and compute the spherical volumes of the UPGMA cones for small n."
]
} |
1704.00492 | 196115574 | Benchmarking methods for 3d hand tracking is still an open problem due to the difficulty of acquiring ground truth data. We introduce a new dataset and benchmarking protocol that is insensitive to the accumulative error of other protocols. To this end, we create testing frame pairs of increasing difficulty and measure the pose estimation error separately for each of them. This approach gives new insights and allows to accurately study the performance of each feature or method without employing a full tracking pipeline. Following this protocol, we evaluate various directional distances in the context of silhouette-based 3d hand tracking, expressed as special cases of a generalized Chamfer distance form. An appropriate parameter setup is proposed for each of them, and a comparative study reveals the best performing method in this context. | Since the earliest days of vision-based hand pose estimation @cite_33 @cite_0 , low-level features like silhouettes @cite_8 , edges @cite_29 , depth @cite_2 , optical flow @cite_8 , shading @cite_4 or a combination of them @cite_22 have been used for hand pose estimation. Although Chamfer distances combined with an edge orientation term have been used in @cite_31 @cite_19 @cite_30 @cite_21 , the different distances have not been thoroughly evaluated for hand pose estimation. While a KD-tree is used in @cite_30 to compute a directional Chamfer distance, Liu al @cite_3 recently proposed a distance transform approach to efficiently use a directional Chamfer distance for shape matching. Different methods of shape matching for pose estimation have been compared in the context of rigid objects @cite_20 or articulated objects @cite_1 . While previous work mainly considered to estimate the pose of a hand in isolation, recent works consider more complicated scenarios where two hands interact with each other @cite_32 @cite_5 or with objects @cite_24 @cite_27 @cite_9 @cite_18 @cite_5 . | {
"cite_N": [
"@cite_30",
"@cite_22",
"@cite_29",
"@cite_3",
"@cite_2",
"@cite_5",
"@cite_20",
"@cite_18",
"@cite_4",
"@cite_8",
"@cite_21",
"@cite_32",
"@cite_19",
"@cite_27",
"@cite_33",
"@cite_9",
"@cite_1",
"@cite_0",
"@cite_24",
"@cite_31"
],
"mid": [
"2122768476",
"2134574424",
"2138672819",
"2011852571",
"2014793423",
"2150457612",
"1538216801",
"2153169563",
"2161604086",
"1649414416",
"2100428290",
"2124419806",
"2162254475",
"2165272793",
"1635989058",
"1995905186",
"2482133267",
"2137940226",
"2543872873",
"2148821982"
],
"abstract": [
"This paper develops probabilistic methods for visual tracking of a three-dimensional geometric hand model from monocular image sequences. We consider a redundant representation in which each model component is described by its position and orientation in the world coordinate frame. A prior model is then defined which enforces the kinematic constraints implied by the model's joints. We show that this prior has a local structure, and is in fact a pairwise Markov random field. Furthermore, our redundant representation allows color and edge-based likelihood measures, such as the Chamfer distance, to be similarly decomposed in cases where there is no self-occlusion. Given this graphical model of hand kinematics, we may track the hand's motion using the recently proposed nonparametric belief propagation (NBP) algorithm. Like particle filters, NBP approximates the posterior distribution over hand configurations as a collection of samples. However, NBP uses the graphical structure to greatly reduce the dimensionality of these distributions, providing improved robustness. Several methods are used to improve NBP's computational efficiency, including a novel KD-tree based method for fast Chamfer distance evaluation. We provide simulations showing that NBP may be used to refine inaccurate model initializations, as well as track hand motion through extended image sequences.",
"We present a model based approach to the integration of multiple cues for tracking high degree of freedom articulated motions and model refinement. We then apply it to the problem of hand tracking using a single camera sequence. Hand tracking is particularly challenging because of occlusions, shading variations, and the high dimensionality of the motion. The novelty of our approach is in the combination of multiple sources of information, which come from edges, optical flow, and shading information in order to refine the model during tracking. We first use a previously formulated generalized version of the gradient-based optical flow constraint, that includes shading flow i.e., the variation of the shading of the object as it rotates with respect to the light source. Using this model we track its complex articulated motion in the presence of shading changes. We use a forward recursive dynamic model to track the motion in response to data derived 3D forces applied to the model. However, due to inaccurate initial shape, the generalized optical flow constraint is violated. We use the error in the generalized optical flow equation to compute generalized forces that correct the model shape at each step. The effectiveness of our approach is demonstrated with experiments on a number of different hand motions with shading changes, rotations and occlusions of significant parts of the hand.",
"In this paper we first describe how we have constructed a 3D deformable Point Distribution Model of the human hand, capturing training data semi-automatically from volume images via a physically-based model. We then show how we have attempted to use this model in tracking an unmarked hand moving with 6 degrees of freedom (plus deformation) in real time using a single video camera. In the course of this we show how to improve on a weighted least-squares pose parameter approximation at little computational cost. We note the successes and shortcomings of our system and discuss how it might be improved.",
"We study the object localization problem in images given a single hand-drawn example or a gallery of shapes as the object model. Although many shape matching algorithms have been proposed for the problem over the decades, chamfer matching remains to be the preferred method when speed and robustness are considered. In this paper, we significantly improve the accuracy of chamfer matching while reducing the computational time from linear to sublinear (shown empirically). Specifically, we incorporate edge orientation information in the matching algorithm such that the resulting cost function is piecewise smooth and the cost variation is tightly bounded. Moreover, we present a sublinear time algorithm for exact computation of the directional chamfer matching score using techniques from 3D distance transforms and directional integral images. In addition, the smooth cost function allows to bound the cost distribution of large neighborhoods and skip the bad hypotheses within. Experiments show that the proposed approach improves the speed of the original chamfer matching upto an order of 45x, and it is much faster than many state of art techniques while the accuracy is comparable.",
"We present a method for automatically estimating the motion of an articulated object filmed by two or more fixed cameras. We focus our work on the case where the quality of the images is poor, and where only an approximation of a geometric model of the tracked object is available. Our technique uses physical forces applied to each rigid part of a kinematic 3D model of the object we are tracking. These forces guide the minimization of the differences between the pose of the 3D model and the pose of the real object in the video images. We use a fast recursive algorithm to solve the dynamical equations of motion of any 3D articulated model. We explain the key parts of our algorithms: how relevant information is extracted from the images, how the forces are created, and how the dynamical equations of motion are solved. A study of what kind of information should be extracted in the images and of when our algorithms fail is also presented. Finally we present some results about the tracking of a person. We also show the application of our method to the tracking of a hand in sequences of images, showing that the kind of information to extract from the images depends on their quality and of the configuration of the cameras.",
"Capturing the motion of two hands interacting with an object is a very challenging task due to the large number of degrees of freedom, self-occlusions, and similarity between the fingers, even in the case of multiple cameras observing the scene. In this paper we propose to use discriminatively learned salient points on the fingers and to estimate the finger-salient point associations simultaneously with the estimation of the hand pose. We introduce a differentiable objective function that also takes edges, optical flow and collisions into account. Our qualitative and quantitative evaluations show that the proposed approach achieves very accurate results for several challenging sequences containing hands and objects in action.",
"In this work, we analyze three different registration algorithms: Chamfer distance matching, the well-known iterated closest points (ICP) and an optic flow based registration. Their pairwise combination is investigated in the context of silhouette based pose estimation. It turns out that Chamfer matching and ICP used in combination do not only perform fairly well with small offset, but also deal with large offset significantly better than the other combinations. We show that by applying different optimized search strategies, the computational cost can be reduced by a factor eight. We further demonstrate the robustness of our method against simultaneous translation and rotation.",
"Due to occlusions, the estimation of the full pose of a human hand interacting with an object is much more challenging than pose recovery of a hand observed in isolation. In this work we formulate an optimization problem whose solution is the 26-DOF hand pose together with the pose and model parameters of the manipulated object. Optimization seeks for the joint hand-object model that (a) best explains the incompleteness of observations resulting from occlusions due to hand-object interaction and (b) is physically plausible in the sense that the hand does not share the same physical space with the object. The proposed method is the first that solves efficiently the continuous, full-DOF, joint hand-object tracking problem based solely on markerless multicamera input. Additionally, it is the first to demonstrate how hand-object interaction can be exploited as a context that facilitates hand pose estimation, instead of being considered as a complicating factor. Extensive quantitative and qualitative experiments with simulated and real world image sequences as well as a comparative evaluation with a state-of-the-art method for pose estimation of isolated hands, support the above findings.",
"A novel model-based approach to 3D hand tracking from monocular video is presented. The 3D hand pose, the hand texture, and the illuminant are dynamically estimated through minimization of an objective function. Derived from an inverse problem formulation, the objective function enables explicit use of temporal texture continuity and shading information while handling important self-occlusions and time-varying illumination. The minimization is done efficiently using a quasi-Newton method, for which we provide a rigorous derivation of the objective function gradient. Particular attention is given to terms related to the change of visibility near self-occlusion boundaries that are neglected in existing formulations. To this end, we introduce new occlusion forces and show that using all gradient terms greatly improves the performance of the method. Qualitative and quantitative experimental results demonstrate the potential of the approach.",
"Sensing of human hand motion is very important for a variety of applications, such as computer animation and athletic performance measurement. Tracking a hand is difficult because the hand has high degree of freedom articulated mechanisms. This paper presents a 3D model-based hand tracking method which is robust to occlusions and local minima. Tracking is performed by minimizing the estimation error of an optical flow and maximizing the overlap between a projected model and a silhouette image. The authors employ stochastic optimization to solve them, which are generally difficult. They present experimental results on tracking from synthetic and real image sequences.",
"This paper sets out a tracking framework, which is applied to the recovery of three-dimensional hand motion from an image sequence. The method handles the issues of initialization, tracking, and recovery in a unified way. In a single input image with no prior information of the hand pose, the algorithm is equivalent to a hierarchical detection scheme, where unlikely pose candidates are rapidly discarded. In image sequences, a dynamic model is used to guide the search and approximate the optimal filtering equations. A dynamic model is given by transition probabilities between regions in parameter space and is learned from training data obtained by capturing articulated motion. The algorithm is evaluated on a number of image sequences, which include hand motion with self-occlusion in front of a cluttered background",
"We propose a method that relies on markerless visual observations to track the full articulation of two hands that interact with each-other in a complex, unconstrained manner. We formulate this as an optimization problem whose 54-dimensional parameter space represents all possible configurations of two hands, each represented as a kinematic structure with 26 Degrees of Freedom (DoFs). To solve this problem, we employ Particle Swarm Optimization (PSO), an evolutionary, stochastic optimization method with the objective of finding the two-hands configuration that best explains observations provided by an RGB-D sensor. To the best of our knowledge, the proposed method is the first to attempt and achieve the articulated motion tracking of two strongly interacting hands. Extensive quantitative and qualitative experiments with simulated and real world image sequences demonstrate that an accurate and efficient solution of this problem is indeed feasible.",
"A method is proposed that can generate a ranked list of plausible three-dimensional hand configurations that best match an input image. Hand pose estimation is formulated as an image database indexing problem, where the closest matches for an input hand image are retrieved from a large database of synthetic hand images. In contrast to previous approaches, the system can function in the presence of clutter, thanks to two novel clutter-tolerant indexing methods. First, a computationally efficient approximation of the image-to-model chamfer distance is obtained by embedding binary edge images into a high-dimensional Euclidean space. Second, a general-purpose, probabilistic line matching method identifies those line segment correspondences between model and input images that are the least likely to have occurred by chance. The performance of this clutter tolerant approach is demonstrated in quantitative experiments with hundreds of real hand images.",
"This paper presents a method for vision based estimation of the pose of human hands in interaction with objects. Despite the fact that most robotics applications of human hand tracking involve grasping and manipulation of objects, the majority of methods in the literature assume a free hand, isolated from the surrounding environment. Our hand tracking method is non-parametric, performing a nearest neighbor search in a large database (100000 entries) of hand poses with and without grasped objects. The system operates in real time, it is robust to self occlusions, object occlusions and segmentation errors, and provides full hand pose reconstruction from markerless video. Temporal consistency in hand pose is taken into account, without explicitly tracking the hand in the high dimensional pose space.",
"Passive sensing of human hand and limb motion is important for a wide range of applications from human-computer interaction to athletic performance measurement. High degree of freedom articulated mechanisms like the human hand are difficult to track because of their large state space and complex image appearance. This article describes a model-based hand tracking system, called DigitEyes, that can recover the state of a 27 DOF hand model from ordinary gray scale images at speeds of up to 10 Hz.",
"In this paper, we propose a prior for hand pose estimation that integrates the direct relation between a manipulating hand and a 3d object. This is of particular interest for a variety of applications since many tasks performed by humans require hand-object interaction. Inspired by the ability of humans to learn the handling of an object from a single example, our focus lies on very sparse training data. We express estimated hand poses in local object coordinates and extract for each individual hand segment, the relative position and orientation as well as contact points on the object. The prior is then modeled as a spatial distribution conditioned to the object. Given a new object of the same object class and new hand dimensions, we can transfer the prior by a procedure involving a geometric warp. In our experiments, we demonstrate that the prior may be used to improve the robustness of a 3d hand tracker and to synthesize a new hand grasping a new object. For this, we integrate the prior into a unified belief propagation framework for tracking and synthesis.",
"In this paper we present a robust and efficient shape matching approach for Marker-less Motion Capture. Extracted features such as contour, gradient orientations and the turning function of the shape are embedded in a 1-D string. We formulate shape matching as a Linear Assignment Problem and propose to use Dynamic Time Warping on the string representation of shapes to discard unlikely correspondences and thereby to reduce ambiguities and spurious local minima. Furthermore, the proposed cost matrix pruning results in robustness to scaling, rotation and topological changes and allows to greatly reduce the computational cost. We show that our approach can track fast human motions where standard articulated Iterative Closest Point algorithms fail.",
"Direct use of the hand as an input device is an attractive method for providing natural human-computer interaction (HCI). Currently, the only technology that satisfies the advanced requirements of hand-based input for HCI is glove-based sensing. This technology, however, has several drawbacks including that it hinders the ease and naturalness with which the user can interact with the computer-controlled environment, and it requires long calibration and setup procedures. Computer vision (CV) has the potential to provide more natural, non-contact solutions. As a result, there have been considerable research efforts to use the hand as an input device for HCI. In particular, two types of research directions have emerged. One is based on gesture classification and aims to extract high-level abstract information corresponding to motion patterns or postures of the hand. The second is based on pose estimation systems and aims to capture the real 3D motion of the hand. This paper presents a literature review on the latter research direction, which is a very challenging problem in the context of HCI.",
"We present a method for tracking a hand while it is interacting with an object. This setting is arguably the one where hand-tracking has most practical relevance, but poses significant additional challenges: strong occlusions by the object as well as self-occlusions are the norm, and classical anatomical constraints need to be softened due to the external forces between hand and object. To achieve robustness to partial occlusions, we use an individual local tracker for each segment of the articulated structure. The segments are connected in a pairwise Markov random field, which enforces the anatomical hand structure through soft constraints on the joints between adjacent segments. The most likely hand configuration is found with belief propagation. Both range and color data are used as input. Experiments are presented for synthetic data with ground truth and for real data of people manipulating objects.",
"This paper compares two methods for object localization from contours: shape context and chamfer matching of templates. In the light of our experiments, we suggest improvements to the shape context: shape contexts are used to find corresponding features between model and image. In real images it is shown that the shape context is highly influenced by clutters; furthermore, even when the object is correctly localized, the feature correspondence may be poor. We show that the robustness of shape matching can be increased by including a figural continuity constraint. The combined shape and continuity cost is minimized using the Viterbi algorithm on features, resulting in improved localization and correspondence. Our algorithm can be generally applied to any feature based shape matching method. Chamfer matching correlates model templates with the distance transform of the edge image. This can be done efficiently using a coarse-to-fine search over the transformation parameters. The method is robust in clutter, however, multiple templates are needed to handle scale, rotation and shape variation. We compare both methods for locating hand shapes in cluttered images, and applied to word recognition in EZ-Gimpy images."
]
} |
1704.00509 | 2951191587 | Recent study shows that a wide deep network can obtain accuracy comparable to a deeper but narrower network. Compared to narrower and deeper networks, wide networks employ relatively less number of layers and have various important benefits, such that they have less running time on parallel computing devices, and they are less affected by gradient vanishing problems. However, the parameter size of a wide network can be very large due to use of large width of each layer in the network. In order to keep the benefits of wide networks meanwhile improve the parameter size and accuracy trade-off of wide networks, we propose a binary tree architecture to truncate architecture of wide networks by reducing the width of the networks. More precisely, in the proposed architecture, the width is continuously reduced from lower layers to higher layers in order to increase the expressive capacity of network with a less increase on parameter size. Also, to ease the gradient vanishing problem, features obtained at different layers are concatenated to form the output of our architecture. By employing the proposed architecture on a baseline wide network, we can construct and train a new network with same depth but considerably less number of parameters. In our experimental analyses, we observe that the proposed architecture enables us to obtain better parameter size and accuracy trade-off compared to baseline networks using various benchmark image classification datasets. The results show that our model can decrease the classification error of baseline from 20.43 to 19.22 on Cifar-100 using only 28 of parameters that baseline has. Code is available at this https URL. | Recently, various architectures of convolutional neural networks (CNNs) have been proposed @cite_1 @cite_4 @cite_25 @cite_18 @cite_21 @cite_6 . @cite_13 , connections between random forest @cite_12 and CNNs are investigated. Inspired by random forest, they embedded routing functions to CNNs and obtain Conditional CNNs. As shown in their experiments, Conditional CNNs with highly branched tree architectures can improve the accuracy-efficiency trade-off. Conditional CNNs can be considered as symmetric full tree architectures. On the other hand, we use an asymmetric tree architecture and concatenate features from different layers. Another related work, a used by FractalNet was proposed in @cite_17 . Fractal architecture can also be considered as a tree architecture. Fractal architecture is different from ours in several aspects. First, the output of a fractal architecture is element-wise mean of features obtained at different layers. Also, all convolutional layers used in fractal architecture use the same width, which may result in a large parameter size if the width is large. | {
"cite_N": [
"@cite_18",
"@cite_4",
"@cite_21",
"@cite_1",
"@cite_6",
"@cite_13",
"@cite_25",
"@cite_12",
"@cite_17"
],
"mid": [
"1686810756",
"",
"2949605076",
"",
"2274287116",
"2290283816",
"2950179405",
"",
"2408279554"
],
"abstract": [
"In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.",
"",
"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 explore 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 on the validation set (3.6 error on the test set) and 17.3 top-1 error on the validation set.",
"",
"Very deep convolutional networks have been central to the largest advances in image recognition performance in recent years. One example is the Inception architecture that has been shown to achieve very good performance at relatively low computational cost. Recently, the introduction of residual connections in conjunction with a more traditional architecture has yielded state-of-the-art performance in the 2015 ILSVRC challenge; its performance was similar to the latest generation Inception-v3 network. This raises the question of whether there are any benefit in combining the Inception architecture with residual connections. Here we give clear empirical evidence that training with residual connections accelerates the training of Inception networks significantly. There is also some evidence of residual Inception networks outperforming similarly expensive Inception networks without residual connections by a thin margin. We also present several new streamlined architectures for both residual and non-residual Inception networks. These variations improve the single-frame recognition performance on the ILSVRC 2012 classification task significantly. We further demonstrate how proper activation scaling stabilizes the training of very wide residual Inception networks. With an ensemble of three residual and one Inception-v4, we achieve 3.08 percent top-5 error on the test set of the ImageNet classification (CLS) challenge",
"This paper investigates the connections between two state of the art classifiers: decision forests (DFs, including decision jungles) and convolutional neural networks (CNNs). Decision forests are computationally efficient thanks to their conditional computation property (computation is confined to only a small region of the tree, the nodes along a single branch). CNNs achieve state of the art accuracy, thanks to their representation learning capabilities. We present a systematic analysis of how to fuse conditional computation with representation learning and achieve a continuum of hybrid models with different ratios of accuracy vs. efficiency. We call this new family of hybrid models conditional networks. Conditional networks can be thought of as: i) decision trees augmented with data transformation operators, or ii) CNNs, with block-diagonal sparse weight matrices, and explicit data routing functions. Experimental validation is performed on the common task of image classification on both the CIFAR and Imagenet datasets. Compared to state of the art CNNs, our hybrid models yield the same accuracy with a fraction of the compute cost and much smaller number of parameters.",
"We propose a deep convolutional neural network architecture codenamed \"Inception\", which was responsible for setting the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC 2014). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. This was achieved by a carefully crafted design that allows for increasing 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 ILSVRC 2014 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.",
"",
"We introduce a design strategy for neural network macro-architecture based on self-similarity. Repeated application of a simple expansion rule generates deep networks whose structural layouts are precisely truncated fractals. These networks contain interacting subpaths of different lengths, but do not include any pass-through or residual connections; every internal signal is transformed by a filter and nonlinearity before being seen by subsequent layers. In experiments, fractal networks match the excellent performance of standard residual networks on both CIFAR and ImageNet classification tasks, thereby demonstrating that residual representations may not be fundamental to the success of extremely deep convolutional neural networks. Rather, the key may be the ability to transition, during training, from effectively shallow to deep. We note similarities with student-teacher behavior and develop drop-path, a natural extension of dropout, to regularize co-adaptation of subpaths in fractal architectures. Such regularization allows extraction of high-performance fixed-depth subnetworks. Additionally, fractal networks exhibit an anytime property: shallow subnetworks provide a quick answer, while deeper subnetworks, with higher latency, provide a more accurate answer."
]
} |
1704.00509 | 2951191587 | Recent study shows that a wide deep network can obtain accuracy comparable to a deeper but narrower network. Compared to narrower and deeper networks, wide networks employ relatively less number of layers and have various important benefits, such that they have less running time on parallel computing devices, and they are less affected by gradient vanishing problems. However, the parameter size of a wide network can be very large due to use of large width of each layer in the network. In order to keep the benefits of wide networks meanwhile improve the parameter size and accuracy trade-off of wide networks, we propose a binary tree architecture to truncate architecture of wide networks by reducing the width of the networks. More precisely, in the proposed architecture, the width is continuously reduced from lower layers to higher layers in order to increase the expressive capacity of network with a less increase on parameter size. Also, to ease the gradient vanishing problem, features obtained at different layers are concatenated to form the output of our architecture. By employing the proposed architecture on a baseline wide network, we can construct and train a new network with same depth but considerably less number of parameters. In our experimental analyses, we observe that the proposed architecture enables us to obtain better parameter size and accuracy trade-off compared to baseline networks using various benchmark image classification datasets. The results show that our model can decrease the classification error of baseline from 20.43 to 19.22 on Cifar-100 using only 28 of parameters that baseline has. Code is available at this https URL. | As shown in @cite_14 @cite_11 , the parameter size of a trained CNN can be reduced by constructing a new CNN with less redundancy of weights. However, these methods may cause a drop on accuracy after a compression of model. With our architecture, we can boost the accuracy with less number of parameters. | {
"cite_N": [
"@cite_14",
"@cite_11"
],
"mid": [
"2950967261",
"1902041153"
],
"abstract": [
"The focus of this paper is speeding up the evaluation of convolutional neural networks. While delivering impressive results across a range of computer vision and machine learning tasks, these networks are computationally demanding, limiting their deployability. Convolutional layers generally consume the bulk of the processing time, and so in this work we present two simple schemes for drastically speeding up these layers. This is achieved by exploiting cross-channel or filter redundancy to construct a low rank basis of filters that are rank-1 in the spatial domain. Our methods are architecture agnostic, and can be easily applied to existing CPU and GPU convolutional frameworks for tuneable speedup performance. We demonstrate this with a real world network designed for scene text character recognition, showing a possible 2.5x speedup with no loss in accuracy, and 4.5x speedup with less than 1 drop in accuracy, still achieving state-of-the-art on standard benchmarks.",
"This paper aims to accelerate the test-time computation of deep convolutional neural networks (CNNs). Unlike existing methods that are designed for approximating linear filters or linear responses, our method takes the nonlinear units into account. We minimize the reconstruction error of the nonlinear responses, subject to a low-rank constraint which helps to reduce the complexity of filters. We develop an effective solution to this constrained nonlinear optimization problem. An algorithm is also presented for reducing the accumulated error when multiple layers are approximated. A whole-model speedup ratio of 4× is demonstrated on a large network trained for ImageNet, while the top-5 error rate is only increased by 0.9 . Our accelerated model has a comparably fast speed as the “AlexNet” [11], but is 4.7 more accurate."
]
} |
1704.00509 | 2951191587 | Recent study shows that a wide deep network can obtain accuracy comparable to a deeper but narrower network. Compared to narrower and deeper networks, wide networks employ relatively less number of layers and have various important benefits, such that they have less running time on parallel computing devices, and they are less affected by gradient vanishing problems. However, the parameter size of a wide network can be very large due to use of large width of each layer in the network. In order to keep the benefits of wide networks meanwhile improve the parameter size and accuracy trade-off of wide networks, we propose a binary tree architecture to truncate architecture of wide networks by reducing the width of the networks. More precisely, in the proposed architecture, the width is continuously reduced from lower layers to higher layers in order to increase the expressive capacity of network with a less increase on parameter size. Also, to ease the gradient vanishing problem, features obtained at different layers are concatenated to form the output of our architecture. By employing the proposed architecture on a baseline wide network, we can construct and train a new network with same depth but considerably less number of parameters. In our experimental analyses, we observe that the proposed architecture enables us to obtain better parameter size and accuracy trade-off compared to baseline networks using various benchmark image classification datasets. The results show that our model can decrease the classification error of baseline from 20.43 to 19.22 on Cifar-100 using only 28 of parameters that baseline has. Code is available at this https URL. | Intuitively, a simpler way to avoid gradient vanishing problem and accelerate the running speed during training and inference is to use a shallow network. @cite_16 , Zagoruyko and Komadakis use a shallow ResNet, and increase the width to make the expressive capacity comparable with narrow-deep ones. Interestingly, they observe that wide-shallow CNNs can outperform its deeper-narrower peer CNNs which have the same parameter size on the Cifar-10 100 classification datasets. On the ILSVRC12 classification task, wide ResNets can also obtain comparable accuracy with smaller depth. Their results draw our attention to consider the wide-shallow DNNs. In their method, the width of each layer is symmetrically increased by the same factor. This will significantly increase the parameter size. Therefore, in this work, the proposed binary architecture truncate architecture of wide networks by reducing width of the networks considering their parameter size and accuracy trade-off. Our motivation for employment of binary tree architecture also considers gradient vanishing problem, which has been shown to be crucial for training DNNs in @cite_7 . | {
"cite_N": [
"@cite_16",
"@cite_7"
],
"mid": [
"2401231614",
"2949892913"
],
"abstract": [
"Deep residual networks were shown to be able to scale up to thousands of layers and still have improving performance. However, each fraction of a percent of improved accuracy costs nearly doubling the number of layers, and so training very deep residual networks has a problem of diminishing feature reuse, which makes these networks very slow to train. To tackle these problems, in this paper we conduct a detailed experimental study on the architecture of ResNet blocks, based on which we propose a novel architecture where we decrease depth and increase width of residual networks. We call the resulting network structures wide residual networks (WRNs) and show that these are far superior over their commonly used thin and very deep counterparts. For example, we demonstrate that even a simple 16-layer-deep wide residual network outperforms in accuracy and efficiency all previous deep residual networks, including thousand-layer-deep networks, achieving new state-of-the-art results on CIFAR, SVHN, COCO, and significant improvements on ImageNet. Our code and models are available at this https URL",
"Very deep convolutional networks with hundreds of layers have led to significant reductions in error on competitive benchmarks. Although the unmatched expressiveness of the many layers can be highly desirable at test time, training very deep networks comes with its own set of challenges. The gradients can vanish, the forward flow often diminishes, and the training time can be painfully slow. To address these problems, we propose stochastic depth, a training procedure that enables the seemingly contradictory setup to train short networks and use deep networks at test time. We start with very deep networks but during training, for each mini-batch, randomly drop a subset of layers and bypass them with the identity function. This simple approach complements the recent success of residual networks. It reduces training time substantially and improves the test error significantly on almost all data sets that we used for evaluation. With stochastic depth we can increase the depth of residual networks even beyond 1200 layers and still yield meaningful improvements in test error (4.91 on CIFAR-10)."
]
} |
1704.00514 | 2950084767 | Keyphrase boundary classification (KBC) is the task of detecting keyphrases in scientific articles and labelling them with respect to predefined types. Although important in practice, this task is so far underexplored, partly due to the lack of labelled data. To overcome this, we explore several auxiliary tasks, including semantic super-sense tagging and identification of multi-word expressions, and cast the task as a multi-task learning problem with deep recurrent neural networks. Our multi-task models perform significantly better than previous state of the art approaches on two scientific KBC datasets, particularly for long keyphrases. | Hard sharing of all hidden layers was introduced in , and popularised in NLP by . Several variants have been introduced, including hard sharing of selected layers @cite_6 and sharing of parts (subspaces) of layers @cite_11 . show that hard parameter sharing is an effective regulariser, also on heterogeneous tasks such as the ones considered here. Hard parameter sharing has been studied for several tasks, including CCG super tagging @cite_6 , text normalisation @cite_7 , neural machine translation @cite_1 @cite_12 , and super-sense tagging @cite_16 . Sharing of information can further be achieved by extending LSTMs with an external memory shared across tasks @cite_13 . A further instance of multi-task learning is to optimise a supervised training objective jointly with an unsupervised training objective, as shown in for natural language generation and auto-encoding, and in for different sequence labelling tasks and language modelling. | {
"cite_N": [
"@cite_7",
"@cite_1",
"@cite_6",
"@cite_16",
"@cite_13",
"@cite_12",
"@cite_11"
],
"mid": [
"2539746012",
"2251743902",
"2516255829",
"2590925815",
"",
"2172589779",
""
],
"abstract": [
"Natural-language processing of historical documents is complicated by the abundance of variant spellings and lack of annotated data. A common approach is to normalize the spelling of historical words to modern forms. We explore the suitability of a deep neural network architecture for this task, particularly a deep bi-LSTM network applied on a character level. Our model compares well to previously established normalization algorithms when evaluated on a diverse set of texts from Early New High German. We show that multi-task learning with additional normalization data can improve our model's performance further.",
"In this paper, we investigate the problem of learning a machine translation model that can simultaneously translate sentences from one source language to multiple target languages. Our solution is inspired by the recently proposed neural machine translation model which generalizes machine translation as a sequence learning problem. We extend the neural machine translation to a multi-task learning framework which shares source language representation and separates the modeling of different target language translation. Our framework can be applied to situations where either large amounts of parallel data or limited parallel data is available. Experiments show that our multi-task learning model is able to achieve significantly higher translation quality over individually learned model in both situations on the data sets publicly available.",
"In all previous work on deep multi-task learning we are aware of, all task supervisions are on the same (outermost) layer. We present a multi-task learning architecture with deep bi-directional RNNs, where different tasks supervision can happen at different layers. We present experiments in syntactic chunking and CCG supertagging, coupled with the additional task of POS-tagging. We show that it is consistently better to have POS supervision at the innermost rather than the outermost layer. We argue that this is because “lowlevel” tasks are better kept at the lower layers, enabling the higher-level tasks to make use of the shared representation of the lower-level tasks. Finally, we also show how this architecture can be used for domain adaptation.",
"Multitask learning has been applied successfully to a range of tasks, mostly morphosyntactic. However, little is known on when MTL works and whether there are data characteristics that help to determine its success. In this paper we evaluate a range of semantic sequence labeling tasks in a MTL setup. We examine different auxiliary tasks, amongst which a novel setup, and correlate their impact to data-dependent conditions. Our results show that MTL is not always effective, significant improvements are obtained only for 1 out of 5 tasks. When successful, auxiliary tasks with compact and more uniform label distributions are preferable.",
"",
"Sequence to sequence learning has recently emerged as a new paradigm in supervised learning. To date, most of its applications focused on only one task and not much work explored this framework for multiple tasks. This paper examines three multi-task learning (MTL) settings for sequence to sequence models: (a) the oneto-many setting - where the encoder is shared between several tasks such as machine translation and syntactic parsing, (b) the many-to-one setting - useful when only the decoder can be shared, as in the case of translation and image caption generation, and (c) the many-to-many setting - where multiple encoders and decoders are shared, which is the case with unsupervised objectives and translation. Our results show that training on a small amount of parsing and image caption data can improve the translation quality between English and German by up to 1.5 BLEU points over strong single-task baselines on the WMT benchmarks. Furthermore, we have established a new state-of-the-art result in constituent parsing with 93.0 F1. Lastly, we reveal interesting properties of the two unsupervised learning objectives, autoencoder and skip-thought, in the MTL context: autoencoder helps less in terms of perplexities but more on BLEU scores compared to skip-thought.",
""
]
} |
1704.00514 | 2950084767 | Keyphrase boundary classification (KBC) is the task of detecting keyphrases in scientific articles and labelling them with respect to predefined types. Although important in practice, this task is so far underexplored, partly due to the lack of labelled data. To overcome this, we explore several auxiliary tasks, including semantic super-sense tagging and identification of multi-word expressions, and cast the task as a multi-task learning problem with deep recurrent neural networks. Our multi-task models perform significantly better than previous state of the art approaches on two scientific KBC datasets, particularly for long keyphrases. | KBC is very similar to named entity recognition (NER), though arguably harder. Deep neural networks have been applied to NER in . Other successful methods rely on conditional random fields, thereby modelling the probability of each output label conditioned on the label at the previous time step. , currently state-of-the-art for NER, stack CRFs on top of recurrent neural networks. We leave exploring such models in combination with multi-task learning for future work. Keyphrase detection methods specific to the scientific domain often use keyphrase gazetteers as features or exploit citation graphs @cite_4 . However, previous methods relied on corpora annotated for type-level identification, not for mention-level identification @cite_5 @cite_2 . While most applications rely on extracting keyphrases (as types), this has the unfortunate consequence that previous work ignores acronyms and other short-hand forms referring to methods, metrics, etc. Further, relying on gazetteers makes overfitting likely, obtaining lower scores on out-of-gazetteer keyphrases. | {
"cite_N": [
"@cite_5",
"@cite_4",
"@cite_2"
],
"mid": [
"2096765155",
"2167329753",
"2566297247"
],
"abstract": [
"Most current statistical natural language processing models use only local features so as to permit dynamic programming in inference, but this makes them unable to fully account for the long distance structure that is prevalent in language use. We show how to solve this dilemma with Gibbs sampling, a simple Monte Carlo method used to perform approximate inference in factored probabilistic models. By using simulated annealing in place of Viterbi decoding in sequence models such as HMMs, CMMs, and CRFs, it is possible to incorporate non-local structure while preserving tractable inference. We use this technique to augment an existing CRF-based information extraction system with long-distance dependency models, enforcing label consistency and extraction template consistency constraints. This technique results in an error reduction of up to 9 over state-of-the-art systems on two established information extraction tasks.",
"While automatic keyphrase extraction has been examined extensively, state-of-theart performance on this task is still much lower than that on many core natural language processing tasks. We present a survey of the state of the art in automatic keyphrase extraction, examining the major sources of errors made by existing systems and discussing the challenges ahead.",
""
]
} |
1704.00552 | 2604593109 | We present the first parser for UCCA, a cross-linguistically applicable framework for semantic representation, which builds on extensive typological work and supports rapid annotation. UCCA poses a challenge for existing parsing techniques, as it exhibits reentrancy (resulting in DAG structures), discontinuous structures and non-terminal nodes corresponding to complex semantic units. To our knowledge, the conjunction of these formal properties is not supported by any existing parser. Our transition-based parser, which uses a novel transition set and features based on bidirectional LSTMs, has value not just for UCCA parsing: its ability to handle more general graph structures can inform the development of parsers for other semantic DAG structures, and in languages that frequently use discontinuous structures. | Recent interest in SDP has yielded numerous works on graph parsing @cite_64 @cite_5 @cite_9 @cite_47 , including tree approximation @cite_6 @cite_45 and joint syntactic semantic parsing @cite_62 @cite_71 . | {
"cite_N": [
"@cite_64",
"@cite_62",
"@cite_9",
"@cite_6",
"@cite_45",
"@cite_71",
"@cite_5",
"@cite_47"
],
"mid": [
"",
"2137132216",
"2130391193",
"2131399212",
"2250381827",
"2467375345",
"2099310262",
"2251237305"
],
"abstract": [
"",
"Current investigations in data-driven models of parsing have shifted from purely syntactic analysis to richer semantic representations, showing that the successful recovery of the meaning of text requires structured analyses of both its grammar and its semantics. In this article, we report on a joint generative history-based model to predict the most likely derivation of a dependency parser for both syntactic and semantic dependencies, in multiple languages. Because these two dependency structures are not isomorphic, we propose a weak synchronization at the level of meaningful subsequences of the two derivations. These synchronized subsequences encompass decisions about the left side of each individual word. We also propose novel derivations for semantic dependency structures, which are appropriate for the relatively unconstrained nature of these graphs. To train a joint model of these synchronized derivations, we make use of a latent variable model of parsing, the Incremental Sigmoid Belief Network ISBN architecture. This architecture induces latent feature representations of the derivations, which are used to discover correlations both within and between the two derivations, providing the first application of ISBNs to a multi-task learning problem. This joint model achieves competitive performance on both syntactic and semantic dependency parsing for several languages. Because of the general nature of the approach, this extension of the ISBN architecture to weakly synchronized syntactic-semantic derivations is also an exemplification of its applicability to other problems where two independent, but related, representations are being learned.",
"As part of the SemEval-2015 shared task on Broad-Coverage Semantic Dependency Parsing, we evaluate the performace of our last year’s system (TurboSemanticParser) on multiple languages and out-of-domain data. Our system is characterized by a feature-rich linear model, that includes scores for first and second-order dependencies (arcs, siblings, grandparents and co-parents). For decoding this second-order model, we solve a linear relaxation of that problem using alternating directions dual decomposition (AD 3 ). The experiments have shown that, even though the parser’s performance in Chinese and Czech attains around 80 (not too far from English performance), domain shift is a serious issue, suggesting domain adaptation as an interesting avenue for future research.",
"We present the Potsdam systems that participated in the semantic dependency parsing shared task of SemEval 2014. They are based on linguistically motivated bidirectional transformations between graphs and trees and on utilization of syntactic dependency parsing. They were entered in both the closed track and the open track of the challenge, recording a peak average labeled F1 score of 78.60.",
"In this shared task paper for SemEval2014 Task 8, we show that most semantic structures can be approximated by trees through a series of almost bijective graph transformations. We transform input graphs, apply off-the-shelf methods from syntactic parsing on the resulting trees, and retrieve output graphs. Using tree approximations, we obtain good results across three semantic formalisms, with a 15.9 error reduction over a stateof-the-art semantic role labeling system on development data. Our system came in 3 6 in the shared task closed track.",
"We present a transition-based parser that jointly produces syntactic and semantic dependencies. It learns a representation of the entire algorithm state, using stack long short-term memories. Our greedy inference algorithm has linear time, including feature extraction. On the CoNLL 2008--9 English shared tasks, we obtain the best published parsing performance among models that jointly learn syntax and semantics.",
"We present an arc-factored statistical model for semantic dependency parsing, as defined by the SemEval 2014 Shared Task 8 on Broad-Coverage Semantic Dependency Parsing. Our entry in the open track placed second in the competition.",
"This paper is a description of our system for SemEval-2015 Task 18: Broad-Coverage Semantic Dependency Parsing. We implement a hybrid parser which benefits from both transition-based and graph-based parsing approaches. In particular, the tree approximation method is explored to take advantage of wellstudied tree parsing techniques. Evaluation on multilingual data sets demonstrates that considerably good semantic analysis can be automatically built by applying state-of-the-art data-driven parsing techniques."
]
} |
1704.00552 | 2604593109 | We present the first parser for UCCA, a cross-linguistically applicable framework for semantic representation, which builds on extensive typological work and supports rapid annotation. UCCA poses a challenge for existing parsing techniques, as it exhibits reentrancy (resulting in DAG structures), discontinuous structures and non-terminal nodes corresponding to complex semantic units. To our knowledge, the conjunction of these formal properties is not supported by any existing parser. Our transition-based parser, which uses a novel transition set and features based on bidirectional LSTMs, has value not just for UCCA parsing: its ability to handle more general graph structures can inform the development of parsers for other semantic DAG structures, and in languages that frequently use discontinuous structures. | Unlike in UCCA, the alignment between AMR concepts and the text is not explicitly marked. While sharing much of this work's motivation, not anchoring the representation in the text complicates the parsing task, as it requires the alignment to be automatically (and imprecisely) detected. Indeed, despite considerable technical effort @cite_19 @cite_13 @cite_2 , concept identification is only about 80 , anchoring allows breaking down sentences into semantically meaningful sub-spans, which is useful for many applications @cite_31 @cite_21 . | {
"cite_N": [
"@cite_21",
"@cite_19",
"@cite_2",
"@cite_31",
"@cite_13"
],
"mid": [
"2949235936",
"2149837184",
"2103575000",
"2108373233",
"2251071803"
],
"abstract": [
"Human evaluation of machine translation normally uses sentence-level measures such as relative ranking or adequacy scales. However, these provide no insight into possible errors, and do not scale well with sentence length. We argue for a semantics-based evaluation, which captures what meaning components are retained in the MT output, thus providing a more fine-grained analysis of translation quality, and enabling the construction and tuning of semantics-based MT. We present a novel human semantic evaluation measure, Human UCCA-based MT Evaluation (HUME), building on the UCCA semantic representation scheme. HUME covers a wider range of semantic phenomena than previous methods and does not rely on semantic annotation of the potentially garbled MT output. We experiment with four language pairs, demonstrating HUME's broad applicability, and report good inter-annotator agreement rates and correlation with human adequacy scores.",
"Abstract Meaning Representation (AMR) is a semantic formalism for which a grow- ing set of annotated examples is avail- able. We introduce the first approach to parse sentences into this representa- tion, providing a strong baseline for fu- ture improvement. The method is based on a novel algorithm for finding a maxi- mum spanning, connected subgraph, em- bedded within a Lagrangian relaxation of an optimization problem that imposes lin- guistically inspired constraints. Our ap- proach is described in the general frame- work of structured prediction, allowing fu- ture incorporation of additional features and constraints, and may extend to other formalisms as well. Our open-source sys- tem, JAMR, is available at: http: github.com jflanigan jamr",
"The Abstract Meaning Representation (AMR) is a representation for open-domain rich semantics, with potential use in fields like event extraction and machine translation. Node generation, typically done using a simple dictionary lookup, is currently an important limiting factor in AMR parsing. We propose a small set of actions that derive AMR subgraphs by transformations on spans of text, which allows for more robust learning of this stage. Our set of construction actions generalize better than the previous approach, and can be learned with a simple classifier. We improve on the previous state-of-the-art result for AMR parsing, boosting end-to-end performance by 3 F @math on both the LDC2013E117 and LDC2014T12 datasets.",
"We reduce phrase-representation parsing to dependency parsing. Our reduction is grounded on a new intermediate representation, \"head-ordered dependency trees\", shown to be isomorphic to constituent trees. By encoding order information in the dependency labels, we show that any off-the-shelf, trainable dependency parser can be used to produce constituents. When this parser is non-projective, we can perform discontinuous parsing in a very natural manner. Despite the simplicity of our approach, experiments show that the resulting parsers are on par with strong baselines, such as the Berkeley parser for English and the best single system in the SPMRL-2014 shared task. Results are particularly striking for discontinuous parsing of German, where we surpass the current state of the art by a wide margin.",
"We align pairs of English sentences and corresponding Abstract Meaning Representations (AMR), at the token level. Such alignments will be useful for downstream extraction of semantic interpretation and generation rules. Our method involves linearizing AMR structures and performing symmetrized EM training. We obtain 86.5 and 83.1 alignment F score on development and test sets."
]
} |
1704.00552 | 2604593109 | We present the first parser for UCCA, a cross-linguistically applicable framework for semantic representation, which builds on extensive typological work and supports rapid annotation. UCCA poses a challenge for existing parsing techniques, as it exhibits reentrancy (resulting in DAG structures), discontinuous structures and non-terminal nodes corresponding to complex semantic units. To our knowledge, the conjunction of these formal properties is not supported by any existing parser. Our transition-based parser, which uses a novel transition set and features based on bidirectional LSTMs, has value not just for UCCA parsing: its ability to handle more general graph structures can inform the development of parsers for other semantic DAG structures, and in languages that frequently use discontinuous structures. | Furthermore, existing transition-based AMR parsers are not general DAG parsers. They are only able to predict a subset of reentrancies and discontinuities, as they may remove nodes before their parents have been predicted @cite_3 . They are thus limited to a sub-class of AMRs in particular, and specifically cannot produce arbitrary DAG parses. 's transition set, on the other hand, allows general DAG parsing. See Appendix for a proof sketch for the completeness of 's transition set. | {
"cite_N": [
"@cite_3"
],
"mid": [
"2515003191"
],
"abstract": [
"Meaning Representation (AMR) is a semantic representation for natural language that embeds annotations related to traditional tasks such as named entity recognition, semantic role labeling, word sense disambiguation and co-reference resolution. We describe a transition-based parser for AMR that parses sentences left-to-right, in linear time. We further propose a test-suite that assesses specific subtasks that are helpful in comparing AMR parsers, and show that our parser is competitive with the state of the art on the LDC2015E86 dataset and that it outperforms state-of-the-art parsers for recovering named entities and handling polarity."
]
} |
1704.00405 | 2613374570 | As for semantic role labeling (SRL) task, when it comes to utilizing parsing information, both traditional methods and recent recurrent neural network (RNN) based methods use the feature engineering way. In this paper, we propose Syntax Aware Long Short Time Memory(SA-LSTM). The structure of SA-LSTM modifies according to dependency parsing information in order to model parsing information directly in an architecture engineering way instead of feature engineering way. We experimentally demonstrate that SA-LSTM gains more improvement from the model architecture. Furthermore, SA-LSTM outperforms the state-of-the-art on CPB 1.0 significantly according to Student t-test ( @math ). | Semantic role labeling (SRL) was first defined by @cite_16 . Early works @cite_16 @cite_7 on SRL got promising result without large annotated SRL corpus. Xue and Palmer built the Chinese Proposition Bank @cite_10 to standardize Chinese SRL research. | {
"cite_N": [
"@cite_16",
"@cite_10",
"@cite_7"
],
"mid": [
"2151170651",
"2063464767",
"99935078"
],
"abstract": [
"We present a system for identifying the semantic relationships, or semantic roles, filled by constituents of a sentence within a semantic frame. Given an input sentence and a target word and frame, the system labels constituents with either abstract semantic roles, such as AGENT or PATIENT, or more domain-specific semantic roles, such as SPEAKER, MESSAGE, and TOPIC.The system is based on statistical classifiers trained on roughly 50,000 sentences that were hand-annotated with semantic roles by the FrameNet semantic labeling project. We then parsed each training sentence into a syntactic tree and extracted various lexical and syntactic features, including the phrase type of each constituent, its grammatical function, and its position in the sentence. These features were combined with knowledge of the predicate verb, noun, or adjective, as well as information such as the prior probabilities of various combinations of semantic roles. We used various lexical clustering algorithms to generalize across possible fillers of roles. Test sentences were parsed, were annotated with these features, and were then passed through the classifiers.Our system achieves 82 accuracy in identifying the semantic role of presegmented constituents. At the more difficult task of simultaneously segmenting constituents and identifying their semantic role, the system achieved 65 precision and 61 recall.Our study also allowed us to compare the usefulness of different features and feature combination methods in the semantic role labeling task. We also explore the integration of role labeling with statistical syntactic parsing and attempt to generalize to predicates unseen in the training data.",
"In this paper, we describe an approach to annotate the propositions in the Penn Chinese Treebank. We describe how diathesis alternation patterns can be used to make coarse sense distinctions for Chinese verbs as a necessary step in annotating the predicate-structure of Chinese verbs. We then discuss the representation scheme we use to label the semantic arguments and adjuncts of the predicates. We discuss several complications for this type of annotation and describe our solutions. We then discuss how a lexical database with predicate-argument structure information can be used to ensure consistent annotation. Finally, we discuss possible applications for this resource.",
"In this paper we address the question of assigning semantic roles to sentences in Chinese. We show that good semantic parsing results for Chinese can be achieved with a small 1100-sentence training set. In order to extract features from Chinese, we describe porting the Collins parser to Chinese, resulting in the best performance currently reported on Chinese syntactic parsing; we include our headrules in the appendix. Finally, we compare English and Chinese semantic-parsing performance. While slight differences in argument labeling make a perfect comparison impossible, our results nonetheless suggest significantly better performance for Chinese. We show that much of this difference is due to grammatical differences between English and Chinese, such as the prevalence of passive in English, and the strict word order constraints on adjuncts in Chinese."
]
} |
1704.00051 | 2604368306 | This paper proposes to tackle open- domain question answering using Wikipedia as the unique knowledge source: the answer to any factoid question is a text span in a Wikipedia article. This task of machine reading at scale combines the challenges of document retrieval (finding the relevant articles) with that of machine comprehension of text (identifying the answer spans from those articles). Our approach combines a search component based on bigram hashing and TF-IDF matching with a multi-layer recurrent neural network model trained to detect answers in Wikipedia paragraphs. Our experiments on multiple existing QA datasets indicate that (1) both modules are highly competitive with respect to existing counterparts and (2) multitask learning using distant supervision on their combination is an effective complete system on this challenging task. | Open-domain QA was originally defined as finding answers in collections of unstructured documents, following the setting of the annual TREC competitions http: trec.nist.gov data qamain.html . With the development of KBs, many recent innovations have occurred in the context of QA from KBs with the creation of resources like WebQuestions @cite_0 and SimpleQuestions @cite_15 based on the Freebase KB @cite_18 , or on automatically extracted KBs, e.g., OpenIE triples and NELL @cite_20 . However, KBs have inherent limitations (incompleteness, fixed schemas) that motivated researchers to return to the original setting of answering from raw text. | {
"cite_N": [
"@cite_0",
"@cite_15",
"@cite_18",
"@cite_20"
],
"mid": [
"2252136820",
"580074167",
"2094728533",
"2090243146"
],
"abstract": [
"In this paper, we train a semantic parser that scales up to Freebase. Instead of relying on annotated logical forms, which is especially expensive to obtain at large scale, we learn from question-answer pairs. The main challenge in this setting is narrowing down the huge number of possible logical predicates for a given question. We tackle this problem in two ways: First, we build a coarse mapping from phrases to predicates using a knowledge base and a large text corpus. Second, we use a bridging operation to generate additional predicates based on neighboring predicates. On the dataset of Cai and Yates (2013), despite not having annotated logical forms, our system outperforms their state-of-the-art parser. Additionally, we collected a more realistic and challenging dataset of question-answer pairs and improves over a natural baseline.",
"Training large-scale question answering systems is complicated because training sources usually cover a small portion of the range of possible questions. This paper studies the impact of multitask and transfer learning for simple question answering; a setting for which the reasoning required to answer is quite easy, as long as one can retrieve the correct evidence given a question, which can be difficult in large-scale conditions. To this end, we introduce a new dataset of 100k questions that we use in conjunction with existing benchmarks. We conduct our study within the framework of Memory Networks (, 2015) because this perspective allows us to eventually scale up to more complex reasoning, and show that Memory Networks can be successfully trained to achieve excellent performance.",
"Freebase is a practical, scalable tuple database used to structure general human knowledge. The data in Freebase is collaboratively created, structured, and maintained. Freebase currently contains more than 125,000,000 tuples, more than 4000 types, and more than 7000 properties. Public read write access to Freebase is allowed through an HTTP-based graph-query API using the Metaweb Query Language (MQL) as a data query and manipulation language. MQL provides an easy-to-use object-oriented interface to the tuple data in Freebase and is designed to facilitate the creation of collaborative, Web-based data-oriented applications.",
"We consider the problem of open-domain question answering (Open QA) over massive knowledge bases (KBs). Existing approaches use either manually curated KBs like Freebase or KBs automatically extracted from unstructured text. In this paper, we present OQA, the first approach to leverage both curated and extracted KBs. A key technical challenge is designing systems that are robust to the high variability in both natural language questions and massive KBs. OQA achieves robustness by decomposing the full Open QA problem into smaller sub-problems including question paraphrasing and query reformulation. OQA solves these sub-problems by mining millions of rules from an unlabeled question corpus and across multiple KBs. OQA then learns to integrate these rules by performing discriminative training on question-answer pairs using a latent-variable structured perceptron algorithm. We evaluate OQA on three benchmark question sets and demonstrate that it achieves up to twice the precision and recall of a state-of-the-art Open QA system."
]
} |
1704.00051 | 2604368306 | This paper proposes to tackle open- domain question answering using Wikipedia as the unique knowledge source: the answer to any factoid question is a text span in a Wikipedia article. This task of machine reading at scale combines the challenges of document retrieval (finding the relevant articles) with that of machine comprehension of text (identifying the answer spans from those articles). Our approach combines a search component based on bigram hashing and TF-IDF matching with a multi-layer recurrent neural network model trained to detect answers in Wikipedia paragraphs. Our experiments on multiple existing QA datasets indicate that (1) both modules are highly competitive with respect to existing counterparts and (2) multitask learning using distant supervision on their combination is an effective complete system on this challenging task. | A second motivation to cast a fresh look at this problem is that of machine comprehension of text, i.e., answering questions after reading a short text or story. That subfield has made considerable progress recently thanks to new deep learning architectures like attention-based and memory-augmented neural networks @cite_3 @cite_14 @cite_29 and release of new training and evaluation datasets like QuizBowl @cite_1 , CNN Daily Mail based on news articles @cite_9 , CBT based on children books @cite_28 , or SQuAD @cite_8 and WikiReading @cite_5 , both based on Wikipedia. An objective of this paper is to test how such new methods can perform in an open-domain QA framework. | {
"cite_N": [
"@cite_14",
"@cite_8",
"@cite_28",
"@cite_9",
"@cite_29",
"@cite_1",
"@cite_3",
"@cite_5"
],
"mid": [
"",
"2427527485",
"2126209950",
"",
"2950527759",
"2130237711",
"2133564696",
"2950663335"
],
"abstract": [
"",
"We present the Stanford Question Answering Dataset (SQuAD), a new reading comprehension dataset consisting of 100,000+ questions posed by crowdworkers on a set of Wikipedia articles, where the answer to each question is a segment of text from the corresponding reading passage. We analyze the dataset to understand the types of reasoning required to answer the questions, leaning heavily on dependency and constituency trees. We build a strong logistic regression model, which achieves an F1 score of 51.0 , a significant improvement over a simple baseline (20 ). However, human performance (86.8 ) is much higher, indicating that the dataset presents a good challenge problem for future research. The dataset is freely available at this https URL",
"We introduce a new test of how well language models capture meaning in children's books. Unlike standard language modelling benchmarks, it distinguishes the task of predicting syntactic function words from that of predicting lower-frequency words, which carry greater semantic content. We compare a range of state-of-the-art models, each with a different way of encoding what has been previously read. We show that models which store explicit representations of long-term contexts outperform state-of-the-art neural language models at predicting semantic content words, although this advantage is not observed for syntactic function words. Interestingly, we find that the amount of text encoded in a single memory representation is highly influential to the performance: there is a sweet-spot, not too big and not too small, between single words and full sentences that allows the most meaningful information in a text to be effectively retained and recalled. Further, the attention over such window-based memories can be trained effectively through self-supervision. We then assess the generality of this principle by applying it to the CNN QA benchmark, which involves identifying named entities in paraphrased summaries of news articles, and achieve state-of-the-art performance.",
"",
"We extend the capabilities of neural networks by coupling them to external memory resources, which they can interact with by attentional processes. The combined system is analogous to a Turing Machine or Von Neumann architecture but is differentiable end-to-end, allowing it to be efficiently trained with gradient descent. Preliminary results demonstrate that Neural Turing Machines can infer simple algorithms such as copying, sorting, and associative recall from input and output examples.",
"Text classification methods for tasks like factoid question answering typically use manually defined string matching rules or bag of words representations. These methods are ineective when question text contains very few individual words (e.g., named entities) that are indicative of the answer. We introduce a recursive neural network (rnn) model that can reason over such input by modeling textual compositionality. We apply our model, qanta, to a dataset of questions from a trivia competition called quiz bowl. Unlike previous rnn models, qanta learns word and phrase-level representations that combine across sentences to reason about entities. The model outperforms multiple baselines and, when combined with information retrieval methods, rivals the best human players.",
"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.",
"We present WikiReading, a large-scale natural language understanding task and publicly-available dataset with 18 million instances. The task is to predict textual values from the structured knowledge base Wikidata by reading the text of the corresponding Wikipedia articles. The task contains a rich variety of challenging classification and extraction sub-tasks, making it well-suited for end-to-end models such as deep neural networks (DNNs). We compare various state-of-the-art DNN-based architectures for document classification, information extraction, and question answering. We find that models supporting a rich answer space, such as word or character sequences, perform best. Our best-performing model, a word-level sequence to sequence model with a mechanism to copy out-of-vocabulary words, obtains an accuracy of 71.8 ."
]
} |
1704.00051 | 2604368306 | This paper proposes to tackle open- domain question answering using Wikipedia as the unique knowledge source: the answer to any factoid question is a text span in a Wikipedia article. This task of machine reading at scale combines the challenges of document retrieval (finding the relevant articles) with that of machine comprehension of text (identifying the answer spans from those articles). Our approach combines a search component based on bigram hashing and TF-IDF matching with a multi-layer recurrent neural network model trained to detect answers in Wikipedia paragraphs. Our experiments on multiple existing QA datasets indicate that (1) both modules are highly competitive with respect to existing counterparts and (2) multitask learning using distant supervision on their combination is an effective complete system on this challenging task. | There are a number of highly developed full pipeline QA approaches using either the Web, as does QuASE @cite_30 , or Wikipedia as a resource, as do Microsoft's AskMSR @cite_2 , IBM's DeepQA @cite_13 and YodaQA @cite_17 @cite_4 --- the latter of which is open source and hence reproducible for comparison purposes. AskMSR is a search-engine based QA system that relies on data redundancy rather than sophisticated linguistic analyses of either questions or candidate answers'', i.e., it does not focus on machine comprehension, as we do. DeepQA is a very sophisticated system that relies on both unstructured information including text documents as well as structured data such as KBs, databases and ontologies to generate candidate answers or vote over evidence. YodaQA is an open source system modeled after DeepQA, similarly combining websites, information extraction, databases and Wikipedia in particular. Our comprehension task is made more challenging by only using a single resource. Comparing against these methods provides a useful datapoint for an upper bound'' benchmark on performance. | {
"cite_N": [
"@cite_30",
"@cite_4",
"@cite_2",
"@cite_13",
"@cite_17"
],
"mid": [
"1646084575",
"1801866228",
"2115758952",
"",
"2607739056"
],
"abstract": [
"Most recent question answering (QA) systems query large-scale knowledge bases (KBs) to answer a question, after parsing and transforming natural language questions to KBs-executable forms (e.g., logical forms). As a well-known fact, KBs are far from complete, so that information required to answer questions may not always exist in KBs. In this paper, we develop a new QA system that mines answers directly from the Web, and meanwhile employs KBs as a significant auxiliary to further boost the QA performance. Specifically, to the best of our knowledge, we make the first attempt to link answer candidates to entities in Freebase, during answer candidate generation. Several remarkable advantages follow: (1) Redundancy among answer candidates is automatically reduced. (2) The types of an answer candidate can be effortlessly determined by those of its corresponding entity in Freebase. (3) Capitalizing on the rich information about entities in Freebase, we can develop semantic features for each answer candidate after linking them to Freebase. Particularly, we construct answer-type related features with two novel probabilistic models, which directly evaluate the appropriateness of an answer candidate's types under a given question. Overall, such semantic features turn out to play significant roles in determining the true answers from the large answer candidate pool. The experimental results show that across two testing datasets, our QA system achieves an 18 54 improvement under F_1 metric, compared with various existing QA systems.",
"We briefly survey the current state of art in the field of Question Answering and present the YodaQA system, an open source framework for this task and a baseline pipeline with reasonable performance. We take a holistic approach, reviewing and aiming to integrate many different question answering task definitions and approaches concerning classes of knowledge bases, question representation and answer generation. To ease performance comparisons of general-purpose QA systems, we also propose an effort in building a new reference QA testing corpus which is a curated and extended version of the TREC corpus.",
"We describe the architecture of the AskMSR question answering system and systematically evaluate contributions of different system components to accuracy. The system differs from most question answering systems in its dependency on data redundancy rather than sophisticated linguistic analyses of either questions or candidate answers. Because a wrong answer is often worse than no answer, we also explore strategies for predicting when the question answering system is likely to give an incorrect answer.",
"",
"This is a preprint, submitted on 2015-03-22. Question Answering as a sub-field of information retrieval and information extraction is recently enjoying renewed pop- ularity, triggered by the publicized success of IBM Watson in the Jeopardy! competition. But Question Answering re- search is now proceeding in several semi-independent tiers depending on the precise task formulation and constraints on the knowledge base, and new researchers entering the field can focus only on various restricted sub-tasks as no modern full-scale software system for QA has been openly available until recently. By our YodaQA system that we introduce here, we seek to re- unite and boost research efforts in Question Answering, pro- viding a modular, open source pipeline for this task — allow- ing integration of various knowledge base paradigms, an- swer production and analysis strategies and using a machine learned models to rank the answers. Within this pipeline, we also supply a baseline QA system inspired by DeepQA with solid performance and propose a reference experimen- tal setup for easy future performance comparisons. In this paper, we review the available open QA platforms, present the architecture of our pipeline, the components of the baseline QA system, and also analyze the system perfor- mance on the reference dataset."
]
} |
1704.00180 | 2572965144 | Many of the state-of-the-art algorithms for gesture recognition are based on Conditional Random Fields (CRFs). Successful approaches, such as the Latent-Dynamic CRFs, extend the CRF by incorporating latent variables, whose values are mapped to the values of the labels. In this paper we propose a novel methodology to set the latent values according to the gesture complexity. We use an heuristic that iterates through the samples associated with each label value, stimating their complexity. We then use it to assign the latent values to the label values. We evaluate our method on the task of recognizing human gestures from video streams. The experiments were performed in binary datasets, generated by grouping different labels. Our results demonstrate that our approach outperforms the arbitrary one in many cases, increasing the accuracy by up to 10 . | Many different probabilistic graphical models have been used for recognizing human gestures and activities, such as: Hidden Markov Models (HMMs) @cite_16 @cite_9 @cite_2 , Dynamic Bayesian Networks (DBNs) @cite_11 , and Conditional Random Fields (CRFs) @cite_3 @cite_1 . These graphical models are either generative or discriminative, depending on whether they make independence assumptions on the input or not. As we usually rely on input that is heavily correlated, discriminative models tend to perform better @cite_0 , especially for sequence labeling tasks @cite_1 @cite_14 . Sminchisescu , for instance, outperformed a HMM approach using a CRF for the task of classifying human motion activities, such as walking and jumping @cite_3 . | {
"cite_N": [
"@cite_14",
"@cite_9",
"@cite_1",
"@cite_3",
"@cite_0",
"@cite_2",
"@cite_16",
"@cite_11"
],
"mid": [
"2117497855",
"262660805",
"2017098439",
"2151214862",
"2163614729",
"",
"2112351484",
"2004671724"
],
"abstract": [
"Many problems in vision involve the prediction of a class label for each frame in an unsegmented sequence. In this paper, we develop a discriminative framework for simultaneous sequence segmentation and labeling which can capture both intrinsic and extrinsic class dynamics. Our approach incorporates hidden state variables which model the sub-structure of a class sequence and learn dynamics between class labels. Each class label has a disjoint set of associated hidden states, which enables efficient training and inference in our model. We evaluated our method on the task of recognizing human gestures from unsegmented video streams and performed experiments on three different datasets of head and eye gestures. Our results demonstrate that our model compares favorably to Support Vector Machines, Hidden Markov Models, and Conditional Random Fields on visual gesture recognition tasks.",
"",
"We present a novel latent discriminative model for human activity recognition. Unlike the approaches that require conditional independence assumptions, our model is very flexible in encoding the full connectivity among observations, latent states, and activity states. The model is able to capture richer class of contextual information in both state-state and observation-state pairs. Although loops are present in the model, we can consider the graphical model as a linear-chain structure, where the exact inference is tractable. Thereby the model is very efficient in both inference and learning. The parameters of the graphical model are learned with the Structured-Support Vector Machine (Structured-SVM). A data-driven approach is used to initialize the latent variables, thereby no hand labeling for the latent states is required. Experimental results on the CAD-120 benchmark dataset show that our model outperforms the state-of-the-art approach by over 5 in both precision and recall, while our model is more efficient in computation.",
"Abstract We describe algorithms for recognizing human motion in monocular video sequences, based on discriminative conditional random fields (CRFs) and maximum entropy Markov models (MEMMs). Existing approaches to this problem typically use generative structures like the hidden Markov model (HMM). Therefore, they have to make simplifying, often unrealistic assumptions on the conditional independence of observations given the motion class labels and cannot accommodate rich overlapping features of the observation or long-term contextual dependencies among observations at multiple timesteps. This makes them prone to myopic failures in recognizing many human motions, because even the transition between simple human activities naturally has temporal segments of ambiguity and overlap. The correct interpretation of these sequences requires more holistic, contextual decisions, where the estimate of an activity at a particular timestep could be constrained by longer windows of observations, prior and even posterior to that timestep. This would not be computationally feasible with a HMM which requires the enumeration of a number of observation sequences exponential in the size of the context window. In this work we follow a different philosophy: instead of restrictively modeling the complex image generation process – the observation, we work with models that can unrestrictedly take it as an input, hence condition on it. Conditional models like the proposed CRFs seamlessly represent contextual dependencies and have computationally attractive properties: they support efficient, exact recognition using dynamic programming, and their parameters can be learned using convex optimization. We introduce conditional graphical models as complementary tools for human motion recognition and present an extensive set of experiments that show not only how these can successfully classify diverse human activities like walking, jumping, running, picking or dancing, but also how they can discriminate among subtle motion styles like normal walks and wander walks.",
"We compare discriminative and generative learning as typified by logistic regression and naive Bayes. We show, contrary to a widely-held belief that discriminative classifiers are almost always to be preferred, that there can often be two distinct regimes of performance as the training set size is increased, one in which each algorithm does better. This stems from the observation—which is borne out in repeated experiments—that while discriminative learning has lower asymptotic error, a generative classifier may also approach its (higher) asymptotic error much faster.",
"",
"A conversation robot that recognizes user's head gestures and uses its results as para-linguistic information is developed. In the conversation, humans exchange linguistic information, which can be obtained by transcription of the utterance, and para-linguistic information, which helps the transmission of linguistic information. Para-linguistic information brings a nuance that cannot be transmitted by linguistic information, and the natural and effective conversation is realized. We recognize user's head gestures as the para-linguistic information in the visual channel. We use the optical flow over the head region as the feature and model them using HMM for the recognition. In actual conversation, while the user performs a gesture, the robot may perform a gesture, too. In this situation, the image sequence captured by the camera mounted on the eyes of the robot includes sways caused by the movement of the camera. To solve this problem, we introduced two artifices. One is for the feature extraction: the optical flow of the body area is used to compensate the swayed images. The other is for the probability models: mode-dependent models are prepared by the MLLR model adaptation technique, and the models are switched according to the motion mode of the robot. Experimental results show the effectiveness of these techniques.",
"It is desirable to know a resident's on-going activities before a robot or a smart system can provide attentive services to meet real human needs. This work addresses the problem of learning and recognizing human daily activities in a dynamic environment. Most currently available approaches learn offline activity models and recognize activities of interest on a real time basis. However, the activity models become outdated when human behaviors or device deployment have changed. It is a tedious and error-prone job to recollect data for retraining the activity models. In such a case, it is important to adapt the learnt activity models to the changes without much human supervision. In this work, we present a self-reconfigurable approach for activity recognition which reconfigures previously learnt activity models and infers multiple activities under a dynamic environment meanwhile pursuing minimal human efforts in relabeling training data by utilizing active-learning assistance."
]
} |
1704.00180 | 2572965144 | Many of the state-of-the-art algorithms for gesture recognition are based on Conditional Random Fields (CRFs). Successful approaches, such as the Latent-Dynamic CRFs, extend the CRF by incorporating latent variables, whose values are mapped to the values of the labels. In this paper we propose a novel methodology to set the latent values according to the gesture complexity. We use an heuristic that iterates through the samples associated with each label value, stimating their complexity. We then use it to assign the latent values to the label values. We evaluate our method on the task of recognizing human gestures from video streams. The experiments were performed in binary datasets, generated by grouping different labels. Our results demonstrate that our approach outperforms the arbitrary one in many cases, increasing the accuracy by up to 10 . | In several approaches, there is an additional layer of latent variables, which are not observed, but inferred. This layer helps to model complex labels @cite_1 @cite_14 @cite_7 @cite_6 @cite_4 and is also used to exploit the underlying semantics between temporal segments @cite_1 . Although the authors agree with the gains of using a layer of latent variables, they have diverging interpretations on how exactly they do so. Hu stated that latent variables represent a subset of a given label. For instance, take an hypothetical class We use the terms class and label value interchangeably. for the activity . The layer of latent variables would help to model the difference between and @cite_1 . On the other hand, Morency claim that latent variables represent different stages of a same label being performed @cite_14 . In a complex gesture or activity, different parts of the motion would have different label values, for example. | {
"cite_N": [
"@cite_14",
"@cite_4",
"@cite_7",
"@cite_1",
"@cite_6"
],
"mid": [
"2117497855",
"2111645492",
"",
"2017098439",
"2106169946"
],
"abstract": [
"Many problems in vision involve the prediction of a class label for each frame in an unsegmented sequence. In this paper, we develop a discriminative framework for simultaneous sequence segmentation and labeling which can capture both intrinsic and extrinsic class dynamics. Our approach incorporates hidden state variables which model the sub-structure of a class sequence and learn dynamics between class labels. Each class label has a disjoint set of associated hidden states, which enables efficient training and inference in our model. We evaluated our method on the task of recognizing human gestures from unsegmented video streams and performed experiments on three different datasets of head and eye gestures. Our results demonstrate that our model compares favorably to Support Vector Machines, Hidden Markov Models, and Conditional Random Fields on visual gesture recognition tasks.",
"Many human action recognition tasks involve data that can be factorized into multiple views such as body postures and hand shapes. These views often interact with each other over time, providing important cues to understanding the action. We present multi-view latent variable discriminative models that jointly learn both view-shared and view-specific sub-structures to capture the interaction between views. Knowledge about the underlying structure of the data is formulated as a multi-chain structured latent conditional model, explicitly learning the interaction between multiple views using disjoint sets of hidden variables in a discriminative manner. The chains are tied using a predetermined topology that repeats over time. We present three topologies — linked, coupled, and linked-coupled — that differ in the type of interaction between views that they model. We evaluate our approach on both segmented and unsegmented human action recognition tasks, using the ArmGesture, the NATOPS, and the ArmGesture-Continuous data. Experimental results show that our approach outperforms previous state-of-the-art action recognition models.",
"",
"We present a novel latent discriminative model for human activity recognition. Unlike the approaches that require conditional independence assumptions, our model is very flexible in encoding the full connectivity among observations, latent states, and activity states. The model is able to capture richer class of contextual information in both state-state and observation-state pairs. Although loops are present in the model, we can consider the graphical model as a linear-chain structure, where the exact inference is tractable. Thereby the model is very efficient in both inference and learning. The parameters of the graphical model are learned with the Structured-Support Vector Machine (Structured-SVM). A data-driven approach is used to initialize the latent variables, thereby no hand labeling for the latent states is required. Experimental results on the CAD-120 benchmark dataset show that our model outperforms the state-of-the-art approach by over 5 in both precision and recall, while our model is more efficient in computation.",
"Automated human activity recognition is an essential task for Human Robot Interaction (HRI). A successful activity recognition system enables an assistant robot to provide precise services. In this paper, we present a two-layered approach that can recognize sub-level activities and high-level activities successively. In the first layer, the low-level activities are recognized based on the RGB-D video. In the second layer, we use the recognized low-level activities as input features for estimating high-level activities. Our model is embedded with a latent node, so that it can capture a richer class of sub-level semantics compared with the traditional approach. Our model is evaluated on a challenging benchmark dataset. We show that the proposed approach outperforms the single-layered approach, suggesting that the hierarchical nature of the model is able to better explain the observed data. The results also show that our model outperforms the state-of-the-art approach in accuracy, precision and recall."
]
} |
1704.00180 | 2572965144 | Many of the state-of-the-art algorithms for gesture recognition are based on Conditional Random Fields (CRFs). Successful approaches, such as the Latent-Dynamic CRFs, extend the CRF by incorporating latent variables, whose values are mapped to the values of the labels. In this paper we propose a novel methodology to set the latent values according to the gesture complexity. We use an heuristic that iterates through the samples associated with each label value, stimating their complexity. We then use it to assign the latent values to the label values. We evaluate our method on the task of recognizing human gestures from video streams. The experiments were performed in binary datasets, generated by grouping different labels. Our results demonstrate that our approach outperforms the arbitrary one in many cases, increasing the accuracy by up to 10 . | Hu and Sung also presented two-layered discriminative models for activity recognition. The model presented by Hu recognizes sub-level activities and high-level activities successively, using the assumption that a high-level activity is composed of multiple sub-activities. While the first layer of the hierarchical model predicts low-level activity for each temporal segment, the second uses the sub-activity to estimate the high-level activity @cite_6 . They added a set of latent variables to enrich the expressiveness of the second layer, and use the sub-activity as an observed variable. The model presented by Sung learns the sub-activities implicitly, by considering them as latent variables @cite_10 . | {
"cite_N": [
"@cite_10",
"@cite_6"
],
"mid": [
"2017695267",
"2106169946"
],
"abstract": [
"Being able to detect and recognize human activities is essential for several applications, including personal assistive robotics. In this paper, we perform detection and recognition of unstructured human activity in unstructured environments. We use a RGBD sensor (Microsoft Kinect) as the input sensor, and compute a set of features based on human pose and motion, as well as based on image and point-cloud information. Our algorithm is based on a hierarchical maximum entropy Markov model (MEMM), which considers a person's activity as composed of a set of sub-activities. We infer the two-layered graph structure using a dynamic programming approach. We test our algorithm on detecting and recognizing twelve different activities performed by four people in different environments, such as a kitchen, a living room, an office, etc., and achieve good performance even when the person was not seen before in the training set.1",
"Automated human activity recognition is an essential task for Human Robot Interaction (HRI). A successful activity recognition system enables an assistant robot to provide precise services. In this paper, we present a two-layered approach that can recognize sub-level activities and high-level activities successively. In the first layer, the low-level activities are recognized based on the RGB-D video. In the second layer, we use the recognized low-level activities as input features for estimating high-level activities. Our model is embedded with a latent node, so that it can capture a richer class of sub-level semantics compared with the traditional approach. Our model is evaluated on a challenging benchmark dataset. We show that the proposed approach outperforms the single-layered approach, suggesting that the hierarchical nature of the model is able to better explain the observed data. The results also show that our model outperforms the state-of-the-art approach in accuracy, precision and recall."
]
} |
1704.00180 | 2572965144 | Many of the state-of-the-art algorithms for gesture recognition are based on Conditional Random Fields (CRFs). Successful approaches, such as the Latent-Dynamic CRFs, extend the CRF by incorporating latent variables, whose values are mapped to the values of the labels. In this paper we propose a novel methodology to set the latent values according to the gesture complexity. We use an heuristic that iterates through the samples associated with each label value, stimating their complexity. We then use it to assign the latent values to the label values. We evaluate our method on the task of recognizing human gestures from video streams. The experiments were performed in binary datasets, generated by grouping different labels. Our results demonstrate that our approach outperforms the arbitrary one in many cases, increasing the accuracy by up to 10 . | In the aforementioned sequential discriminative models where the latent values have a direct mapping to the label, the assignment of the number of latent values per label value is chosen arbitrarily. Typically the same number of latent values is given for all label values @cite_4 @cite_1 @cite_14 . Our work proposes a systematic way to distribute these, presenting significant performance gains in the tested datasets. Our method is valid for all the models of this category, but we mainly build upon the LDCRFs proposed by Morency @cite_14 , since they were used to measure the performance improvements achieved by adopting our approach. | {
"cite_N": [
"@cite_1",
"@cite_14",
"@cite_4"
],
"mid": [
"2017098439",
"2117497855",
"2111645492"
],
"abstract": [
"We present a novel latent discriminative model for human activity recognition. Unlike the approaches that require conditional independence assumptions, our model is very flexible in encoding the full connectivity among observations, latent states, and activity states. The model is able to capture richer class of contextual information in both state-state and observation-state pairs. Although loops are present in the model, we can consider the graphical model as a linear-chain structure, where the exact inference is tractable. Thereby the model is very efficient in both inference and learning. The parameters of the graphical model are learned with the Structured-Support Vector Machine (Structured-SVM). A data-driven approach is used to initialize the latent variables, thereby no hand labeling for the latent states is required. Experimental results on the CAD-120 benchmark dataset show that our model outperforms the state-of-the-art approach by over 5 in both precision and recall, while our model is more efficient in computation.",
"Many problems in vision involve the prediction of a class label for each frame in an unsegmented sequence. In this paper, we develop a discriminative framework for simultaneous sequence segmentation and labeling which can capture both intrinsic and extrinsic class dynamics. Our approach incorporates hidden state variables which model the sub-structure of a class sequence and learn dynamics between class labels. Each class label has a disjoint set of associated hidden states, which enables efficient training and inference in our model. We evaluated our method on the task of recognizing human gestures from unsegmented video streams and performed experiments on three different datasets of head and eye gestures. Our results demonstrate that our model compares favorably to Support Vector Machines, Hidden Markov Models, and Conditional Random Fields on visual gesture recognition tasks.",
"Many human action recognition tasks involve data that can be factorized into multiple views such as body postures and hand shapes. These views often interact with each other over time, providing important cues to understanding the action. We present multi-view latent variable discriminative models that jointly learn both view-shared and view-specific sub-structures to capture the interaction between views. Knowledge about the underlying structure of the data is formulated as a multi-chain structured latent conditional model, explicitly learning the interaction between multiple views using disjoint sets of hidden variables in a discriminative manner. The chains are tied using a predetermined topology that repeats over time. We present three topologies — linked, coupled, and linked-coupled — that differ in the type of interaction between views that they model. We evaluate our approach on both segmented and unsegmented human action recognition tasks, using the ArmGesture, the NATOPS, and the ArmGesture-Continuous data. Experimental results show that our approach outperforms previous state-of-the-art action recognition models."
]
} |
1704.00327 | 2604284492 | This paper addresses the problem of designing a trajectory tracking control law for a quadrotor UAV, subsequent to complete failure of a single rotor. The control design problem considers the reduced state space which excludes the angular velocity and orientation about the vertical body axis. The proposed controller enables the quadrotor to track the orientation of this axis, and consequently any prescribed position trajectory using only three rotors. The control design is carried out in two stages. First, in order to track the reduced attitude dynamics, a geometric controller with two input torques is designed on the Lie-Group @math . This is then extended to @math by designing a saturation based feedback law, in order to track the center of mass position with bounded thrust. The control law for the complete dynamics achieves exponential tracking for all initial conditions lying in an open-dense subset. The novelty of the geometric control design is in its ability to effectively execute aggressive, global maneuvers despite complete loss of a rotor. Numerical simulations on models of a variable pitch and a conventional quadrotor have been presented to demonstrate the practical applicability of the control design. | In recent times, globally stabilizing geometric controllers which exploit the intrinsic structure of the underlying manifold, have been developed. Here, singularities due to attitude parameterizations or input-output decoupling are avoided. Approaches such as that in @cite_29 and @cite_11 stabilize mechanical systems on Lie Groups using nonlinear proportional-derivative (PD) control. One such pioneering control design for quadrotors on the Lie Group @math has been presented in @cite_9 and @cite_19 . Reduced attitude stabilization to a fixed point on @math with two control torques is presented in @cite_6 . In @cite_12 , the primary axis of a rigid body on @math as well as the angular velocity about this axis, are tracked using three independent torques. In @cite_3 , global reduced attitude tracking with three torques is achieved by constructing a synergistic family of potential functions on @math . To the best of our knowledge, none of the above mentioned control laws are suitable for trajectory tracking with three functioning rotors (i.e. two torque inputs and a net thrust). | {
"cite_N": [
"@cite_9",
"@cite_29",
"@cite_6",
"@cite_3",
"@cite_19",
"@cite_12",
"@cite_11"
],
"mid": [
"2109656638",
"2124228020",
"1545855501",
"2110636401",
"",
"2149841884",
"57077127"
],
"abstract": [
"This paper provides new results for the tracking control of a quadrotor unmanned aerial vehicle (UAV). The UAV has four input degrees of freedom, namely the magnitudes of the four rotor thrusts, that are used to control the six translational and rotational degrees of freedom, and to achieve asymptotic tracking of four outputs, namely, three position variables for the vehicle center of mass and the direction of one vehicle body-fixed axis. A globally defined model of the quadrotor UAV rigid body dynamics is introduced as a basis for the analysis. A nonlinear tracking controller is developed on the special Euclidean group SE(3) and it is shown to have desirable closed loop properties that are almost global. Several numerical examples, including an example in which the quadrotor recovers from being initially upside down, illustrate the versatility of the controller.",
"We present a general intrinsic tracking controller design for fully-actuated simple mechanical systems, when the configuration space is one of a general class of Lie groups. We first express a state-feedback controller in terms of a function-the \"error function\"-satisfying certain regularity conditions. If an error function can be found, then a general smooth and bounded reference trajectory may be tracked asymptotically from almost every initial condition, with locally exponential convergence. Asymptotic convergence from almost every initial condition is referred to as \"almost-global\" asymptotic stability. Error functions may be shown to exist on any compact Lie group, or any Lie group diffeomorphic to the product of a compact Lie group and R sup n . This covers many cases of practical interest, such as SO(n), SE(n), their subgroups, and direct products. We show here that for compact Lie groups the dynamic configuration-feedback controller obtained by composing the full state-feedback law with an exponentially convergent velocity observer is also almost-globally asymptotically stable with respect to the tracking error. We emphasize that no invariance is needed for these results. However, for the special case where the kinetic energy is left-invariant, we show that the explicit expression of these controllers does not require coordinates on the Lie group. The controller constructions are demonstrated on SO(3), and simulated for the axi-symmetric top. Results show excellent performance.",
"This paper focuses on a new geometric approach to (fully actuated) control systems on the sphere. Our control laws exploit the basic and intuitive notions of geodesic direction and of distance between points, and generalize the classical proportional plus derivative feedback (PD) without the need of arbitrary local coordinate charts. The stability analysis relies on an appropriate Lyapunov function, where the notion of distance and its properties are exploited. This methodology then applies to spin-axis stabilization of a spacecraft actuated by only two control torques: discarding the rotation about the unactuated axis, a reduced system is considered, whose state is in fact defined on the sphere. For this reduced stabilization problem our approach allows us not only to deal optimally with the inevitable singularity, but also to achieve simplicity, versatility and (coordinate independent) adaptive capabilities.",
"Abstract In this paper, we propose a hybrid feedback based on a “synergistic” potential function that achieves global asymptotic stabilization of a desired orientation on the n -sphere with a nominal robustness to measurement disturbances, a task that is not possible by classical feedback–be it smooth, nonsmooth, periodic, or any combination thereof–due to the topological structure of the sphere. We extend this basic result to a tracking controller for the reduced attitude–or pointing direction–of a rigid body and provide a method to remove jumps in the controlled torque by backstepping. The proposed hybrid feedback is compared with a similar smooth feedback in simulation, where it is illustrated that the hybrid feedback overcomes performance limitations inherent to the smooth feedback. We provide two examples of a synergistic potential function–one defined on a general sphere and the other on the unit circle.",
"",
"The control task of tracking a reference pointing direction (the attitude about the pointing direction is irrelevant) while obtaining a desired angular velocity (PDAV) around the pointing direction using geometric techniques is addressed here. Existing geometric controllers developed on the two-sphere only address the tracking of a reference pointing direction while driving the angular velocity about the pointing direction to zero. In this paper a tracking controller on the two-sphere, able to address the PDAV control task, is developed globally in a geometric frame work, to avoid problems related to other attitude representations such as unwinding (quaternions) or singularities (Euler angles). An attitude error function is constructed resulting in a control system with desired tracking performance for rotational maneuvers with large initial attitude angular velocity errors and the ability to negotiate bounded modeling inaccuracies. The tracking ability of the developed control system is evaluated by comparing its performance with an existing geometric controller on the two-sphere and by numerical simulations, showing improved performance for large initial attitude errors, smooth transitions between desired angular velocities and the ability to negotiate bounded modeling inaccuracies.",
"This talk will outline a comprehensive set of modeling, analysis and design techniques for a class of mechanical systems. We concern ourselves with simple mechanical control systems, that is, systems whose Lagrangian is kinetic energy minus potential energy. Example devices include robotic manipulators, aerospace and underwater vehicles, and mechanisms that locomote exploiting nonholonomic constraints. Borrowing techniques from nonlinear control and geometric mechanics, we propose a coordinateinvariant control theory for this class of systems. First, we take a Riemannian geometric approach to modeling systems dened on smooth manifolds, subject to nonholonomic constraints, external forces and control forces. We also model mechanical systems on groups and symmetries. Second, we analyze some control-theoretic properties of this class of systems, including controllability, averaged response to oscillatory controls, and kinematic reductions. Finally, we exploit the modeling and analysis results to tackle control design problems. Starting from controllability and kinematic reduction assumptions we propose some algorithms for generating and tracking trajectories."
]
} |
1704.00077 | 2593363143 | This paper proposes a geodesic-distance-based feature that encodes global information for improved video segmentation algorithms. The feature is a joint histogram of intensity and geodesic distances, where the geodesic distances are computed as the shortest paths between superpixels via their boundaries. We also incorporate adaptive voting weights and spatial pyramid configurations to include spatial information into the geodesic histogram feature and show that this further improves results. The feature is generic and can be used as part of various algorithms. In experiments, we test the geodesic histogram feature by incorporating it into two existing video segmentation frameworks. This leads to significantly better performance in 3D video segmentation benchmarks on two datasets. | Many video segmentation works propose diverse features to capture various kinds of information in order to estimate the similarity between the components of the video. Appearance can be represented by features based on color @cite_12 @cite_7 , texture @cite_25 , and soft boundaries @cite_1 . Motion related features have also been utilized often, including short-term motion features based on optical flow @cite_15 @cite_9 and long-term motion features based on trajectories @cite_13 @cite_8 @cite_26 @cite_19 . Superpixel shape is used to compute the similarities among superpixels across frames @cite_7 . Some works discuss the choice of features to use @cite_20 as well as the method to incorporate various kinds of features into affinity matrices @cite_17 . | {
"cite_N": [
"@cite_26",
"@cite_7",
"@cite_8",
"@cite_9",
"@cite_1",
"@cite_19",
"@cite_15",
"@cite_20",
"@cite_13",
"@cite_25",
"@cite_12",
"@cite_17"
],
"mid": [
"2028204211",
"2099814188",
"2068994826",
"2460260369",
"1994356125",
"",
"2165745222",
"1500853963",
"1496571393",
"1484228140",
"2030346542",
"2024938892"
],
"abstract": [
"As early stage of video processing, we introduce an iterative trajectory merging algorithm that produces a region-based and hierarchical representation of the video sequence, called the Trajectory Binary Partition Tree (BPT). From this representation, many analysis and graph cut techniques can be used to extract partitions or objects that are useful in the context of specific applications. In order to define trajectories and to create a precise merging algorithm, color and motion cues have to be used. Both types of informations are very useful to characterize objects but present strong differences of behavior in the spatial and the temporal dimensions. On the one hand, scenes and objects are rich in their spatial color distributions, but these distributions are rather stable over time. Object motion, on the other hand, presents simple structures and low spatial variability but may change from frame to frame. The proposed algorithm takes into account this key difference and relies on different models and associated metrics to deal with color and motion information. We show that the proposed algorithm outperforms existing hierarchical video segmentation algorithms and provides more stable and precise regions.",
"Unsupervised video segmentation is a challenging problem because it involves a large amount of data, and image segments undergo noisy variations in color, texture and motion with time. However, there are significant redundancies that can help disambiguate the effects of noise. To exploit these redundancies and obtain the most spatio-temporally consistent video segmentation, we formulate the problem as a consistent labeling problem by exploiting higher order image structure. A label stands for a specific moving segment. Each segment (or region) is treated as a random variable which is to be assigned a label. Regions assigned the same label comprise a 3D space-time segment, or a region tube. The labels can also be automatically created or terminated at any frame in the video sequence, to allow objects entering or leaving the scene. To formulate this problem, we use the CRF (conditional random field) model. Unlike conventional CRF which has only unary and binary potentials, we also use higher order potentials to favor label consistency among disconnected spatial and temporal segments. Compared to region tracking based methods, the main advantages of the proposed algorithm are two fold: (1) the label consistency constraints are imposed on multiple regions but in a soft manner, and (2) the labeling decision is postponed until the confidence in the labeling is high. We compare our results with a recent state-of-the-art video segmentation algorithm and show that our results are quantitatively and qualitatively better.",
"Video provides not only rich visual cues such as motion and appearance, but also much less explored long-range temporal interactions among objects. We aim to capture such interactions and to construct a powerful intermediate-level video representation for subsequent recognition. Motivated by this goal, we seek to obtain spatio-temporal oversegmentation of a video into regions that respect object boundaries and, at the same time, associate object pixels over many video frames. The contributions of this paper are two-fold. First, we develop an efficient spatiotemporal video segmentation algorithm, which naturally incorporates long-range motion cues from the past and future frames in the form of clusters of point tracks with coherent motion. Second, we devise a new track clustering cost function that includes occlusion reasoning, in the form of depth ordering constraints, as well as motion similarity along the tracks. We evaluate the proposed approach on a challenging set of video sequences of office scenes from feature length movies.",
"Video object segmentation is challenging due to fast moving objects, deforming shapes, and cluttered backgrounds. Optical flow can be used to propagate an object segmentation over time but, unfortunately, flow is often inaccurate, particularly around object boundaries. Such boundaries are precisely where we want our segmentation to be accurate. To obtain accurate segmentation across time, we propose an efficient algorithm that considers video segmentation and optical flow estimation simultaneously. For video segmentation, we formulate a principled, multiscale, spatio-temporal objective function that uses optical flow to propagate information between frames. For optical flow estimation, particularly at object boundaries, we compute the flow independently in the segmented regions and recompose the results. We call the process object flow and demonstrate the effectiveness of jointly optimizing optical flow and video segmentation using an iterative scheme. Experiments on the SegTrack v2 and Youtube-Objects datasets show that the proposed algorithm performs favorably against the other state-of-the-art methods.",
"Computational and memory costs restrict spectral techniques to rather small graphs, which is a serious limitation especially in video segmentation. In this paper, we propose the use of a reduced graph based on superpixels. In contrast to previous work, the reduced graph is reweighted such that the resulting segmentation is equivalent, under certain assumptions, to that of the full graph. We consider equivalence in terms of the normalized cut and of its spectral clustering relaxation. The proposed method reduces runtime and memory consumption and yields on par results in image and video segmentation. Further, it enables an efficient data representation and update for a new streaming video segmentation approach that also achieves state-of-the-art performance.",
"",
"We propose a novel model for the spatio-temporal clustering of trajectories based on motion, which applies to challenging street-view video sequences of pedestrians captured by a mobile camera. A key contribution of our work is the introduction of novel probabilistic region trajectories, motivated by the non-repeatability of segmentation of frames in a video sequence. Hierarchical image segments are obtained by using a state-of-the-art hierarchical segmentation algorithm, and connected from adjacent frames in a directed acyclic graph. The region trajectories and measures of confidence are extracted from this graph using a dynamic programming-based optimisation. Our second main contribution is a Bayesian framework with a twofold goal: to learn the optimal, in a maximum likelihood sense, Random Forests classifier of motion patterns based on video features, and construct a unique graph from region trajectories of different frames, lengths and hierarchical levels. Finally, we demonstrate the use of Isomap for effective spatio-temporal clustering of the region trajectories of pedestrians. We support our claims with experimental results on new and existing challenging video sequences.",
"Due to its importance, video segmentation has regained interest recently. However, there is no common agreement about the necessary ingredients for best performance. This work contributes a thorough analysis of various within- and between-frame affinities suitable for video segmentation. Our results show that a frame-based superpixel segmentation combined with a few motion and appearance-based affinities are sufficient to obtain good video segmentation performance. A second contribution of the paper is the extension of [1] to include motion-cues, which makes the algorithm globally aware of motion, thus improving its performance for video sequences. Finally, we contribute an extension of an established image segmentation benchmark [1] to videos, allowing coarse-to-fine video segmentations and multiple human annotations. Our results are tested on BMDS [2], and compared to existing methods.",
"Unsupervised learning requires a grouping step that defines which data belong together. A natural way of grouping in images is the segmentation of objects or parts of objects. While pure bottom-up segmentation from static cues is well known to be ambiguous at the object level, the story changes as soon as objects move. In this paper, we present a method that uses long term point trajectories based on dense optical flow. Defining pair-wise distances between these trajectories allows to cluster them, which results in temporally consistent segmentations of moving objects in a video shot. In contrast to multi-body factorization, points and even whole objects may appear or disappear during the shot. We provide a benchmark dataset and an evaluation method for this so far uncovered setting.",
"We study the recognition of surfaces made from different materials such as concrete, rug, marble, or leather on the basis of their textural appearance. Such natural textures arise from spatial variation of two surface attributes: (1) reflectance and (2) surface normal. In this paper, we provide a unified model to address both these aspects of natural texture. The main idea is to construct a vocabulary of prototype tiny surface patches with associated local geometric and photometric properties. We call these 3D textons. Examples might be ridges, grooves, spots or stripes or combinations thereof. Associated with each texton is an appearance vector, which characterizes the local irradiance distribution, represented as a set of linear Gaussian derivative filter outputs, under different lighting and viewing conditions. Given a large collection of images of different materials, a clustering approach is used to acquire a small (on the order of 100) 3D texton vocabulary. Given a few (1 to 4) images of any material, it can be characterized using these textons. We demonstrate the application of this representation for recognition of the material viewed under novel lighting and viewing conditions. We also illustrate how the 3D texton model can be used to predict the appearance of materials under novel conditions.",
"We present an efficient and scalable technique for spatiotemporal segmentation of long video sequences using a hierarchical graph-based algorithm. We begin by over-segmenting a volumetric video graph into space-time regions grouped by appearance. We then construct a “region graph” over the obtained segmentation and iteratively repeat this process over multiple levels to create a tree of spatio-temporal segmentations. This hierarchical approach generates high quality segmentations, which are temporally coherent with stable region boundaries, and allows subsequent applications to choose from varying levels of granularity. We further improve segmentation quality by using dense optical flow to guide temporal connections in the initial graph. We also propose two novel approaches to improve the scalability of our technique: (a) a parallel out-of-core algorithm that can process volumes much larger than an in-core algorithm, and (b) a clip-based processing algorithm that divides the video into overlapping clips in time, and segments them successively while enforcing consistency. We demonstrate hierarchical segmentations on video shots as long as 40 seconds, and even support a streaming mode for arbitrarily long videos, albeit without the ability to process them hierarchically.",
"Video segmentation has become an important and active research area with a large diversity of proposed approaches. Graph-based methods, enabling top-performance on recent benchmarks, consist of three essential components: 1. powerful features account for object appearance and motion similarities; 2. spatio-temporal neighborhoods of pixels or superpixels (the graph edges) are modeled using a combination of those features; 3. video segmentation is formulated as a graph partitioning problem. While a wide variety of features have been explored and various graph partition algorithms have been proposed, there is surprisingly little research on how to construct a graph to obtain the best video segmentation performance. This is the focus of our paper. We propose to combine features by means of a classifier, use calibrated classifier outputs as edge weights and define the graph topology by edge selection. By learning the graph (without changes to the graph partitioning method), we improve the results of the best performing video segmentation algorithm by 6 on the challenging VSB100 benchmark, while reducing its runtime by 55 , as the learnt graph is much sparser."
]
} |
1704.00077 | 2593363143 | This paper proposes a geodesic-distance-based feature that encodes global information for improved video segmentation algorithms. The feature is a joint histogram of intensity and geodesic distances, where the geodesic distances are computed as the shortest paths between superpixels via their boundaries. We also incorporate adaptive voting weights and spatial pyramid configurations to include spatial information into the geodesic histogram feature and show that this further improves results. The feature is generic and can be used as part of various algorithms. In experiments, we test the geodesic histogram feature by incorporating it into two existing video segmentation frameworks. This leads to significantly better performance in 3D video segmentation benchmarks on two datasets. | Geodesic distances provide appearance-based similarity estimates. Geodesic distances have been applied widely on segmentation related problems on images @cite_23 @cite_3 @cite_11 . A feature based on geodesic distance for matching images of deformed objects has been introduced in @cite_3 . The authors showed that the geodesic distance could be invariant to object deformations, by encoding pixels as color histograms on the surrounding pixels that have the same geodesic distances. The geodesic distance is also used to propose object segments on images @cite_23 , which is based on the correlation between the object boundary and the change in the geodesic distance transform. Several video segmentation methods have employed geodesic distance for various purposes. The salient object segmentation framework uses a geodesic distance in each frame to estimate the objectness of superpixels @cite_18 on a per frame basis. Further work further proposes a spatio-temporal geodesic distance @cite_11 that extends image segmentation to video segmentation. However, the proposed spatio-temporal distance has to be constrained to be temporally non-decreasing to preserve the metric property, thus limiting the robustness of the method. | {
"cite_N": [
"@cite_18",
"@cite_3",
"@cite_23",
"@cite_11"
],
"mid": [
"261873710",
"2166835585",
"",
"2134839354"
],
"abstract": [
"We present an approach for identifying a set of candidate objects in a given image. This set of candidates can be used for object recognition, segmentation, and other object-based image parsing tasks. To generate the proposals, we identify critical level sets in geodesic distance transforms computed for seeds placed in the image. The seeds are placed by specially trained classifiers that are optimized to discover objects. Experiments demonstrate that the presented approach achieves significantly higher accuracy than alternative approaches, at a fraction of the computational cost.",
"We propose a novel framework to build descriptors of local intensity that are invariant to general deformations. In this framework, an image is embedded as a 2D surface in 3D space, with intensity weighted relative to distance in x-y. We show that as this weight increases, geodesic distances on the embedded surface are less affected by image deformations. In the limit, distances are deformation invariant. We use geodesic sampling to get neighborhood samples for interest points, and then use a geodesic-intensity histogram (GIH) as a deformation invariant local descriptor. In addition to its invariance, the new descriptor automatically finds its support region. This means it can safely gather information from a large neighborhood to improve discriminability. Furthermore, we propose a matching method for this descriptor that is invariant to affine lighting changes. We have tested this new descriptor on interest point matching for two data sets, one with synthetic deformation and lighting change, and another with real non-affine deformations. Our method shows promising matching results compared to several other approaches",
"",
"An interactive framework for soft segmentation and matting of natural images and videos is presented in this paper. The proposed technique is based on the optimal, linear time, computation of weighted geodesic distances to the user-provided scribbles, from which the whole data is automatically segmented. The weights are based on spatial and or temporal gradients, without explicit optical flow or any advanced and often computationally expensive feature detectors. These could be naturally added to the proposed framework as well if desired, in the form of weights in the geodesic distances. A localized refinement step follows this fast segmentation in order to accurately compute the corresponding matte function. Additional constraints into the distance definition permit to efficiently handle occlusions such as people or objects crossing each other in a video sequence. The presentation of the framework is complemented with numerous and diverse examples, including extraction of moving foreground from dynamic background, and comparisons with the recent literature."
]
} |
1703.10847 | 2746068898 | Most existing neural network models for music generation use recurrent neural networks. However, the recent WaveNet model proposed by DeepMind shows that convolutional neural networks (CNNs) can also generate realistic musical waveforms in the audio domain. Following this light, we investigate using CNNs for generating melody (a series of MIDI notes) one bar after another in the symbolic domain. In addition to the generator, we use a discriminator to learn the distributions of melodies, making it a generative adversarial network (GAN). Moreover, we propose a novel conditional mechanism to exploit available prior knowledge, so that the model can generate melodies either from scratch, by following a chord sequence, or by conditioning on the melody of previous bars (e.g. a priming melody), among other possibilities. The resulting model, named MidiNet, can be expanded to generate music with multiple MIDI channels (i.e. tracks). We conduct a user study to compare the melody of eight-bar long generated by MidiNet and by Google's MelodyRNN models, each time using the same priming melody. Result shows that MidiNet performs comparably with MelodyRNN models in being realistic and pleasant to listen to, yet MidiNet's melodies are reported to be much more interesting. | A large number of deep neural network models have been proposed lately for music generation. This includes models for generating a melody sequence or audio waveforms by following a few priming notes @cite_17 @cite_31 @cite_13 @cite_5 @cite_3 @cite_22 , accompanying a melody sequence with music of other parts @cite_33 , or playing a duet with human @cite_1 @cite_8 . | {
"cite_N": [
"@cite_22",
"@cite_33",
"@cite_8",
"@cite_1",
"@cite_3",
"@cite_5",
"@cite_31",
"@cite_13",
"@cite_17"
],
"mid": [
"2559246505",
"2560316200",
"",
"2561662441",
"2953331651",
"2951535099",
"",
"",
""
],
"abstract": [
"This paper presents an efficient implementation of the Wavenet generation process called Fast Wavenet. Compared to a naive implementation that has complexity O(2^L) (L denotes the number of layers in the network), our proposed approach removes redundant convolution operations by caching previous calculations, thereby reducing the complexity to O(L) time. Timing experiments show significant advantages of our fast implementation over a naive one. While this method is presented for Wavenet, the same scheme can be applied anytime one wants to perform autoregressive generation or online prediction using a model with dilated convolution layers. The code for our method is publicly available.",
"This paper introduces DeepBach, a graphical model aimed at modeling polyphonic music and specifically hymn-like pieces. We claim that, after being trained on the chorale harmonizations by Johann Sebastian Bach, our model is capable of generating highly convincing chorales in the style of Bach. DeepBach's strength comes from the use of pseudo-Gibbs sampling coupled with an adapted representation of musical data. This is in contrast with many automatic music composition approaches which tend to compose music sequentially. Our model is also steerable in the sense that a user can constrain the generation by imposing positional constraints such as notes, rhythms or cadences in the generated score. We also provide a plugin on top of the MuseScore music editor making the interaction with DeepBach easy to use.",
"",
"Several methods exist for a computer to generate music based on data including Markov chains, recurrent neural networks, recombinancy, and grammars. We explore the use of unit selection and concatenation as a means of generating music using a procedure based on ranking, where, we consider a unit to be a variable length number of measures of music. We first examine whether a unit selection method, that is restricted to a finite size unit library, can be sufficient for encompassing a wide spectrum of music. We do this by developing a deep autoencoder that encodes a musical input and reconstructs the input by selecting from the library. We then describe a generative model that combines a deep structured semantic model (DSSM) with an LSTM to predict the next unit, where units consist of four, two, and one measures of music. We evaluate the generative model using objective metrics including mean rank and accuracy and with a subjective listening test in which expert musicians are asked to complete a forced-choiced ranking task. We compare our model to a note-level generative baseline that consists of a stacked LSTM trained to predict forward by one note.",
"In this paper we propose a novel model for unconditional audio generation based on generating one audio sample at a time. We show that our model, which profits from combining memory-less modules, namely autoregressive multilayer perceptrons, and stateful recurrent neural networks in a hierarchical structure is able to capture underlying sources of variations in the temporal sequences over very long time spans, on three datasets of different nature. Human evaluation on the generated samples indicate that our model is preferred over competing models. We also show how each component of the model contributes to the exhibited performance.",
"Generative models in vision have seen rapid progress due to algorithmic improvements and the availability of high-quality image datasets. In this paper, we offer contributions in both these areas to enable similar progress in audio modeling. First, we detail a powerful new WaveNet-style autoencoder model that conditions an autoregressive decoder on temporal codes learned from the raw audio waveform. Second, we introduce NSynth, a large-scale and high-quality dataset of musical notes that is an order of magnitude larger than comparable public datasets. Using NSynth, we demonstrate improved qualitative and quantitative performance of the WaveNet autoencoder over a well-tuned spectral autoencoder baseline. Finally, we show that the model learns a manifold of embeddings that allows for morphing between instruments, meaningfully interpolating in timbre to create new types of sounds that are realistic and expressive.",
"",
"",
""
]
} |
1703.10847 | 2746068898 | Most existing neural network models for music generation use recurrent neural networks. However, the recent WaveNet model proposed by DeepMind shows that convolutional neural networks (CNNs) can also generate realistic musical waveforms in the audio domain. Following this light, we investigate using CNNs for generating melody (a series of MIDI notes) one bar after another in the symbolic domain. In addition to the generator, we use a discriminator to learn the distributions of melodies, making it a generative adversarial network (GAN). Moreover, we propose a novel conditional mechanism to exploit available prior knowledge, so that the model can generate melodies either from scratch, by following a chord sequence, or by conditioning on the melody of previous bars (e.g. a priming melody), among other possibilities. The resulting model, named MidiNet, can be expanded to generate music with multiple MIDI channels (i.e. tracks). We conduct a user study to compare the melody of eight-bar long generated by MidiNet and by Google's MelodyRNN models, each time using the same priming melody. Result shows that MidiNet performs comparably with MelodyRNN models in being realistic and pleasant to listen to, yet MidiNet's melodies are reported to be much more interesting. | Song from PI @cite_6 is a hierarchical RNN model that uses a hierarchy of recurrent layers to generate not only the melody but also the drums and chords, leading to a multi-track pop song. This model nicely demonstrates the ability of RNNs in generating multiple sequences simultaneously. However, it requires prior knowledge of the musical scale and some profiles of the melody to be generated @cite_6 , which is not needed in many other models, including MidiNet. | {
"cite_N": [
"@cite_6"
],
"mid": [
"2953100410"
],
"abstract": [
"We present a novel framework for generating pop music. Our model is a hierarchical Recurrent Neural Network, where the layers and the structure of the hierarchy encode our prior knowledge about how pop music is composed. In particular, the bottom layers generate the melody, while the higher levels produce the drums and chords. We conduct several human studies that show strong preference of our generated music over that produced by the recent method by Google. We additionally show two applications of our framework: neural dancing and karaoke, as well as neural story singing."
]
} |
1703.10847 | 2746068898 | Most existing neural network models for music generation use recurrent neural networks. However, the recent WaveNet model proposed by DeepMind shows that convolutional neural networks (CNNs) can also generate realistic musical waveforms in the audio domain. Following this light, we investigate using CNNs for generating melody (a series of MIDI notes) one bar after another in the symbolic domain. In addition to the generator, we use a discriminator to learn the distributions of melodies, making it a generative adversarial network (GAN). Moreover, we propose a novel conditional mechanism to exploit available prior knowledge, so that the model can generate melodies either from scratch, by following a chord sequence, or by conditioning on the melody of previous bars (e.g. a priming melody), among other possibilities. The resulting model, named MidiNet, can be expanded to generate music with multiple MIDI channels (i.e. tracks). We conduct a user study to compare the melody of eight-bar long generated by MidiNet and by Google's MelodyRNN models, each time using the same priming melody. Result shows that MidiNet performs comparably with MelodyRNN models in being realistic and pleasant to listen to, yet MidiNet's melodies are reported to be much more interesting. | DeepBach @cite_33 , proposed by Sony CSL, is specifically designed for composing polyphonic four-part chorale music in the style of J. S. Bach. It is an RNN-based model that allows enforcing user-defined constraints such as rhythm, notes, parts, chords and cadences. | {
"cite_N": [
"@cite_33"
],
"mid": [
"2560316200"
],
"abstract": [
"This paper introduces DeepBach, a graphical model aimed at modeling polyphonic music and specifically hymn-like pieces. We claim that, after being trained on the chorale harmonizations by Johann Sebastian Bach, our model is capable of generating highly convincing chorales in the style of Bach. DeepBach's strength comes from the use of pseudo-Gibbs sampling coupled with an adapted representation of musical data. This is in contrast with many automatic music composition approaches which tend to compose music sequentially. Our model is also steerable in the sense that a user can constrain the generation by imposing positional constraints such as notes, rhythms or cadences in the generated score. We also provide a plugin on top of the MuseScore music editor making the interaction with DeepBach easy to use."
]
} |
1703.10717 | 2605195953 | We propose a new equilibrium enforcing method paired with a loss derived from the Wasserstein distance for training auto-encoder based Generative Adversarial Networks. This method balances the generator and discriminator during training. Additionally, it provides a new approximate convergence measure, fast and stable training and high visual quality. We also derive a way of controlling the trade-off between image diversity and visual quality. We focus on the image generation task, setting a new milestone in visual quality, even at higher resolutions. This is achieved while using a relatively simple model architecture and a standard training procedure. | Deep Convolutional GANs @cite_1 (DCGANs) first introduced a convolutional architecture which led to improved visual quality. More recently, Energy Based GANs @cite_16 (EBGANs) were proposed as a class of GANs that aims to model the discriminator @math as an energy function. This variant converges more stably and is both easy to train and robust to hyper-parameter variations. The authors attribute some of these benefits to the larger number of targets in the discriminator. EBGAN likewise implements its discriminator as an auto-encoder with a per-pixel error. | {
"cite_N": [
"@cite_16",
"@cite_1"
],
"mid": [
"2521028896",
"2173520492"
],
"abstract": [
"We introduce the \"Energy-based Generative Adversarial Network\" model (EBGAN) which views the discriminator as an energy function that attributes low energies to the regions near the data manifold and higher energies to other regions. Similar to the probabilistic GANs, a generator is seen as being trained to produce contrastive samples with minimal energies, while the discriminator is trained to assign high energies to these generated samples. Viewing the discriminator as an energy function allows to use a wide variety of architectures and loss functionals in addition to the usual binary classifier with logistic output. Among them, we show one instantiation of EBGAN framework as using an auto-encoder architecture, with the energy being the reconstruction error, in place of the discriminator. We show that this form of EBGAN exhibits more stable behavior than regular GANs during training. We also show that a single-scale architecture can be trained to generate high-resolution images.",
"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."
]
} |
1703.10977 | 2604741677 | We present fully formalized proofs of some central theorems from combinatorics. These are Dilworth's decomposition theorem, Mirsky's theorem, Hall's marriage theorem and the Erd o s-Szekeres theorem. Dilworth's decomposition theorem is the key result among these. It states that in any finite partially ordered set (poset), the size of a smallest chain cover and a largest antichain are the same. Mirsky's theorem is a dual of Dilworth's decomposition theorem, which states that in any finite poset, the size of a smallest antichain cover and a largest chain are the same. We use Dilworth's theorem in the proofs of Hall's Marriage theorem and the Erd o s-Szekeres theorem. The combinatorial objects involved in these theorems are sets and sequences. All the proofs are formalized in the Coq proof assistant. We develop a library of definitions and facts that can be used as a framework for formalizing other theorems on finite posets. | Rudnicki @cite_14 presents a formalization of Dilworth's decomposition theorem in Mizar. In the same paper they also provide a proof of the Erd o s-Szekeres theorem using Dilworth's theorem. A separate proof of the Hall's marriage theorem in Mizar appeared in @cite_7 . Jiang and Nipkow @cite_9 also presented two different proofs of Hall's theorem in Isabelle HOL. We have used a different theorem prover and formalized all of these results in a single framework. Our work is closest to the work of @cite_14 . However, we added extra results (Hall's theorem) in the same framework. The proof we mechanize for Hall's theorem uses Dilworth's theorem and we formalize Hall's theorem in both of its popular forms. The first form deals with the matching in a bipartite graph and the second form is about sequence of distinct representatives (SDR) for a collection of finite sets. We also provide a clear compilation of some useful results on finite sets and posets that can be used for mechanizing other important results from the combinatorics of finite structures. | {
"cite_N": [
"@cite_9",
"@cite_14",
"@cite_7"
],
"mid": [
"2098909266",
"2166565522",
""
],
"abstract": [
"In this paper I will discuss the fundamental ideas behind proof assistants: What are they and what is a proof anyway? I give a short history of the main ideas, emphasizing the way they ensure the correctness of the mathematics formalized. I will also briefly discuss the places where proof assistants are used and how we envision their extended use in the future. While being an introduction into the world of proof assistants and the main issues behind them, this paper is also a position paper that pushes the further use of proof assistants. We believe that these systems will become the future of mathematics, where definitions, statements, computations and proofs are all available in a computerized form. An important application is and will be in computer supported modelling and verification of systems. But there is still a long road ahead and I will indicate what we believe is needed for the further proliferation of proof assistants.",
"The following theorem is due to Dilworth [8]: Let P be a partially ordered set. If the maximal number of elements in an independent subset (anti-chain) of P is k, then P is the union of k chains (cliques). In this article we formalize an elegant proof of the above theorem for finite posets by Perles [13]. The result is then used in proving the case of infinite posets following the original proof of Dilworth [8]. A dual of Dilworth’s theorem also holds: a poset with maximum clique m is a union of m independent sets. The proof of this dual fact is considerably easier; we follow the proof by Mirsky [11]. Mirsky states also a corollary that a poset of r × s + 1 elements possesses a clique of size r + 1 or an independent set of size s + 1, or both. This corollary is then used to prove the result of Erdős and Szekeres [9]. Instead of using posets, we drop reflexivity and state the facts about antisymmetric and transitive relations.",
""
]
} |
1703.10757 | 2604548710 | We proposed a deep learning method for interpretable diabetic retinopathy (DR) detection. The visual-interpretable feature of the proposed method is achieved by adding the regression activation map (RAM) after the global averaging pooling layer of the convolutional networks (CNN). With RAM, the proposed model can localize the discriminative regions of an retina image to show the specific region of interest in terms of its severity level. We believe this advantage of the proposed deep learning model is highly desired for DR detection because in practice, users are not only interested with high prediction performance, but also keen to understand the insights of DR detection and why the adopted learning model works. In the experiments conducted on a large scale of retina image dataset, we show that the proposed CNN model can achieve high performance on DR detection compared with the state-of-the-art while achieving the merits of providing the RAM to highlight the salient regions of the input image. | The most work most related to our method is @cite_3 in which class activation map is proposed to characterize the weighted activation maps after global average pooling or global maximum pooling layer. This idea has recently been generalized to time series analysis to localize the significant regions in the raw data @cite_17 . In this paper, we extend the method @cite_3 from a classification to a regression setting and shed light on DR detection problem. | {
"cite_N": [
"@cite_3",
"@cite_17"
],
"mid": [
"2950328304",
"2951213053"
],
"abstract": [
"In this work, we revisit the global average pooling layer proposed in [13], and shed light on how it explicitly enables the convolutional neural network to have remarkable localization ability despite being trained on image-level labels. While this technique was previously proposed as a means for regularizing training, we find that it actually builds a generic localizable deep representation that can be applied to a variety of tasks. Despite the apparent simplicity of global average pooling, we are able to achieve 37.1 top-5 error for object localization on ILSVRC 2014, which is remarkably close to the 34.2 top-5 error achieved by a fully supervised CNN approach. We demonstrate that our network is able to localize the discriminative image regions on a variety of tasks despite not being trained for them",
"We propose a simple but strong baseline for time series classification from scratch with deep neural networks. Our proposed baseline models are pure end-to-end without any heavy preprocessing on the raw data or feature crafting. The proposed Fully Convolutional Network (FCN) achieves premium performance to other state-of-the-art approaches and our exploration of the very deep neural networks with the ResNet structure is also competitive. The global average pooling in our convolutional model enables the exploitation of the Class Activation Map (CAM) to find out the contributing region in the raw data for the specific labels. Our models provides a simple choice for the real world application and a good starting point for the future research. An overall analysis is provided to discuss the generalization capability of our models, learned features, network structures and the classification semantics."
]
} |
1703.10399 | 2518196255 | The integrity of messages in vehicular ad-hoc networks has been extensively studied by the research community, resulting in the IEEE~1609.2 standard, which provides typical integrity guarantees. However, the correctness of message contents is still one of the main challenges of applying dependable and secure vehicular ad-hoc networks. One important use case is the validity of position information contained in messages: position verification mechanisms have been proposed in the literature to provide this functionality. A more general approach to validate such information is by applying misbehavior detection mechanisms. In this paper, we consider misbehavior detection by enhancing two position verification mechanisms and fusing their results in a generalized framework using subjective logic. We conduct extensive simulations using VEINS to study the impact of traffic density, as well as several types of attackers and fractions of attackers on our mechanisms. The obtained results show the proposed framework can validate position information as effectively as existing approaches in the literature, without tailoring the framework specifically for this use case. | Leinmüller et al. @cite_1 proposed a number of different position verification detectors. This variety was in part the inspiration for our work, where one important contribution is the ability to incorporate different information sources. We now describe one of their mechanisms in detail, which we aim to improve: the Acceptance Range Threshold (ART). This mechanism essentially relies on the fact that transmission range is limited. Therefore, if attackers manipulate their position to be at some distance away from their actual position, some vehicles will receive messages with position information outside of their reception range. In their work, the authors assume a fixed reception range of 250 meters, and their attacker is stationary. In our work, we propose to improve their work by estimating the reception range with more accuracy, and by adapting the mechanisms' output from a binary value (legitimate or falsified) to a subjective logic opinion (see Section ). This opinion can then be fused with other detection results. | {
"cite_N": [
"@cite_1"
],
"mid": [
"2073137959"
],
"abstract": [
"Inter-vehicle communication is regarded as one of the major applications of mobile ad hoc networks (MANETs). Compared to MANETs or wireless sensor networks (WSNs), these so-called vehicular ad hoc networks (VANETs) have unique requirements on network protocols. The requirements result mainly from node mobility and the demands of position-dependent applications. On the routing layer, those requirements are well met by geographic routing protocols. Functional research on geographic routing has already reached a considerable level, whereas security aspects have only been recently taken into account. Position information dissemination has been identified as being crucial for geographic routing since forged position information has severe impact regarding both performance and security. In this work, we first summarize the problems that arise from falsified position data. We then propose a framework that contains different detection mechanisms in order to mitigate or lessen these problems. Our developed mechanisms are capable of recognizing nodes cheating about their position in beacons (periodic position dissemination in most single-path geographic routing protocols, e.g., GPSR). Unlike other proposals described in the literature, our detection system does not rely on additional hardware or special nodes, which would contradict the ad hoc approach. Instead, we use a number of different independent sensors to quickly give an estimation of the trustworthiness of other nodes' position claims. The different sensors run either autonomously on every single node, or they require cooperation between neighboring nodes. The simulation evaluation proves that the combination of autonomous and cooperative position verification mechanisms successfully discloses most nodes disseminating false position information, and thereby widely prevents attacks using position cheating."
]
} |
1703.10399 | 2518196255 | The integrity of messages in vehicular ad-hoc networks has been extensively studied by the research community, resulting in the IEEE~1609.2 standard, which provides typical integrity guarantees. However, the correctness of message contents is still one of the main challenges of applying dependable and secure vehicular ad-hoc networks. One important use case is the validity of position information contained in messages: position verification mechanisms have been proposed in the literature to provide this functionality. A more general approach to validate such information is by applying misbehavior detection mechanisms. In this paper, we consider misbehavior detection by enhancing two position verification mechanisms and fusing their results in a generalized framework using subjective logic. We conduct extensive simulations using VEINS to study the impact of traffic density, as well as several types of attackers and fractions of attackers on our mechanisms. The obtained results show the proposed framework can validate position information as effectively as existing approaches in the literature, without tailoring the framework specifically for this use case. | Raya et al. @cite_8 have described a framework to combine various data-centric detection mechanisms. However, unlike our approach, their main focus is computing node trustworthiness, which is then used to evaluate the actual belief in the received message. This approach is fundamentally based on trust evaluation, which can be done using different logic frameworks, such as Dempster-Shafer theory or Bayesian inference. One of their results is that when uncertainty is high, Dempster-Shafer theory performs well; we use subjective logic in our work, which is an improvement over Dempster-Shafer theory @cite_6 . We also go beyond their work conceptually, building on earlier work by Dietzel et al. @cite_9 , and represent detection results in the logic framework, rather than just trust between nodes. This makes our work more flexible: it can conceptually represent aggregated information and can be tuned depending on network parameters, as discussed in earlier work @cite_10 . | {
"cite_N": [
"@cite_9",
"@cite_10",
"@cite_6",
"@cite_8"
],
"mid": [
"1963914923",
"2248544492",
"2110689325",
"2114902045"
],
"abstract": [
"Vehicle to Vehicle (V2V) and Vehicle to Infrastruc- ture (V2I) communication aims to increase safety, efficiency, and comfort of drivers. Vehicles periodically broadcast their current status, such as position, velocity, and other information. Received information is stored in a local knowledge base, often called world model, and used for application decisions. Because of the potential impact, V2V communication is an interesting target for malicious attackers. Message integrity protection using cryptographic signatures only protects against outsider attackers. In addition to signatures, misbehavior detection mechanisms comparable to intrusion detection systems (IDS) are needed to detect insider attackers. Given the complexity and large number of foreseen V2V and V2I applications, misbehavior detection cannot be a one-size-fits-all solution. In this paper, we present a flexible framework that can combine a range of different misbehavior detection mechanisms by modeling their outputs using subjective logic. We demonstrate the feasibility of our framework by using a combination of existing detection mechanisms to increase their misbehavior detection results.",
"Information dissemination is one of the most-discussed applications for vehicular ad hoc networks (VANETs) and other ad hoc networks. To provide dependability for applications, information dissemination must be resilient against different kinds of attacks. Especially insider attackers, which may create valid messages that cannot easily be detected using cryptographic signatures alone, pose a viable threat to information dependability. Many proposals in existing work offer solutions to detect individual attack patterns using data consistency checks and other means. We propose a generic framework that can integrate a wide range of existing detection mechanisms, allows to combine their outputs to improve attack detection, and enables mechanism adaptation based on current attack likelihood. We employ subjective logic opinions, which enable flexible security mechanism output representation, and which we extend to support continuing operation in dynamic networks, such as VANETs. Simulation results show that our framework improves detection accuracy compared to applying individual mechanisms.",
"We first describe a metric for uncertain probabilities called opinion, and subsequently a set of logical operators that can be used for logical reasoning with uncertain propositions. This framework which is called subjective logic uses elements from the Dempster-Shafer belief theory and we show that it is compatible with binary logic and probability calculus.",
"We argue that the traditional notion of trust as a relation among entities, while useful, becomes insufficient for emerging data-centric mobile ad hoc networks. In these systems, setting the data trust level equal to the trust level of the data- providing entity would ignore system salient features, rendering applications ineffective and systems inflexible. This would be even more so if their operation is ephemeral, i.e., characterized by short-lived associations in volatile environments. In this paper, we address this challenge by extending the traditional notion of trust to data-centric trust: trustworthiness attributed to node-reported data per se. We propose a framework for data-centric trust establishment: First, trust in each individual piece of data is computed; then multiple, related but possibly contradictory, data are combined; finally, their validity is inferred by a decision component based on one of several evidence evaluation techniques. We consider and evaluate an instantiation of our framework in vehicular networks as a case study. Our simulation results show that our scheme is highly resilient to attackers and converges stably to the correct decision."
]
} |
1703.10399 | 2518196255 | The integrity of messages in vehicular ad-hoc networks has been extensively studied by the research community, resulting in the IEEE~1609.2 standard, which provides typical integrity guarantees. However, the correctness of message contents is still one of the main challenges of applying dependable and secure vehicular ad-hoc networks. One important use case is the validity of position information contained in messages: position verification mechanisms have been proposed in the literature to provide this functionality. A more general approach to validate such information is by applying misbehavior detection mechanisms. In this paper, we consider misbehavior detection by enhancing two position verification mechanisms and fusing their results in a generalized framework using subjective logic. We conduct extensive simulations using VEINS to study the impact of traffic density, as well as several types of attackers and fractions of attackers on our mechanisms. The obtained results show the proposed framework can validate position information as effectively as existing approaches in the literature, without tailoring the framework specifically for this use case. | Stübing et al. @cite_11 have proposed a different approach; rather than developing a generic framework for misbehavior detection, they have developed a framework to combine several information sources that are all concerned with correctness of position and movement information. Specifically, their approach combines several autonomous data-centric mechanisms (path prediction and maneuver recognition), which allow them to accurately predict the movement of neighboring vehicles. However, their work has two main disadvantages: it cannot detect certain types of attacks (e.g., when the attacker consistently falsifies her position by a fixed vector), and it cannot be extended or combined to work with other detectors without further work. In our framework, it is possible to integrate their detection results, and the concrete detectors we improve in this paper can detect exactly the attack that the framework by Stübing et al. @cite_11 cannot. | {
"cite_N": [
"@cite_11"
],
"mid": [
"1980772307"
],
"abstract": [
"Security is a prerequisite for successfully establishing Car-to-X technology on the market. Sender authentication, message integrity as well as plausibility of the message content represent one of the key aspects in this domain. While the first two requirements are covered by means of cryptography, the latter is still subject to current research. Recent works in the area of vehicle behavior analysis include promising approaches, but are still prone to errors in case of highly dynamic driving maneuvers. In this work we present a novel two-stage verification process for reliable Car-to-X mobility data verification. The first stage consists of a mobility estimator realized by a Kalman filter. The Kalman filter is thereby used to evaluate received mobility data with respect to the path history of vehicles. In the second stage a plausibility check for highly dynamic traffic situations is applied using probabilistic traffic maneuver recognition based on Hidden Markov Models. The overall framework is implemented and its effectiveness is evaluated by means of real world experiments."
]
} |
1703.10399 | 2518196255 | The integrity of messages in vehicular ad-hoc networks has been extensively studied by the research community, resulting in the IEEE~1609.2 standard, which provides typical integrity guarantees. However, the correctness of message contents is still one of the main challenges of applying dependable and secure vehicular ad-hoc networks. One important use case is the validity of position information contained in messages: position verification mechanisms have been proposed in the literature to provide this functionality. A more general approach to validate such information is by applying misbehavior detection mechanisms. In this paper, we consider misbehavior detection by enhancing two position verification mechanisms and fusing their results in a generalized framework using subjective logic. We conduct extensive simulations using VEINS to study the impact of traffic density, as well as several types of attackers and fractions of attackers on our mechanisms. The obtained results show the proposed framework can validate position information as effectively as existing approaches in the literature, without tailoring the framework specifically for this use case. | Subjective logic @cite_6 is a framework for probabilistic information fusion, which is capable of representing not just a probabilistic truth value, but also a measure of uncertainty. It is similar to the more well-known Dempster-Shafer Theory, with the advantage that it integrates uncertainty directly, rather than adding it as a separate component. This has the advantage that fusion becomes easier. Subjective logic expresses the truth value of a statement through so-called opinions @math , which consist of a belief, disbelief, uncertainty and base rate. An intuitive interpretation is that belief is the probability that the statement is true, disbelief is the probability that it is false, and uncertainty represents the confidence in this evaluation. The base rate expresses the probability in the absence of evidence, which in this paper is assumed to be @math . An opinion can be converted into a prediction by computing the expectation, @math . Opinions are usually held by subjects about objects: for example, several detection mechanisms (subjects) can have different opinions about an object (a new packet). Subjective logic provides operators to fuse the opinions of these detectors with certain constraints. | {
"cite_N": [
"@cite_6"
],
"mid": [
"2110689325"
],
"abstract": [
"We first describe a metric for uncertain probabilities called opinion, and subsequently a set of logical operators that can be used for logical reasoning with uncertain propositions. This framework which is called subjective logic uses elements from the Dempster-Shafer belief theory and we show that it is compatible with binary logic and probability calculus."
]
} |
1703.10344 | 2054434916 | Wikipedia entity pages are a valuable source of information for direct consumption and for knowledge-base construction, update and maintenance. Facts in these entity pages are typically supported by references. Recent studies show that as much as 20 of the references are from online news sources. However, many entity pages are incomplete even if relevant information is already available in existing news articles. Even for the already present references, there is often a delay between the news article publication time and the reference time. In this work, we therefore look at Wikipedia through the lens of news and propose a novel news-article suggestion task to improve news coverage in Wikipedia, and reduce the lag of newsworthy references. Our work finds direct application, as a precursor, to Wikipedia page generation and knowledge-base acceleration tasks that rely on relevant and high quality input sources. We propose a two-stage supervised approach for suggesting news articles to entity pages for a given state of Wikipedia. First, we suggest news articles to Wikipedia entities (article-entity placement) relying on a rich set of features which take into account the salience and relative authority of entities, and the novelty of news articles to entity pages. Second, we determine the exact section in the entity page for the input article (article-section placement) guided by class-based section templates. We perform an extensive evaluation of our approach based on ground-truth data that is extracted from external references in Wikipedia. We achieve a high precision value of up to 93 in the article-entity suggestion stage and upto 84 for the article-section placement. Finally, we compare our approach against competitive baselines and show significant improvements. | In this task, given specific information extraction templates, a given corpus is analyzed in order to find worthwhile mentions of an entity or snippets that match the templates. Balog @cite_5 @cite_0 recommend news citations for an entity. Prior to that, the news articles are classified for their appropriateness for an entity, where as features for the classification task they use entity, document, entity-document and temporal features. The best performing features are those that measure similarity between an entity and the news document. @cite_10 consider the problem of knowledge base completion, through question answering and complete missing facts in Freebase based on templates, i.e. bornIn . | {
"cite_N": [
"@cite_0",
"@cite_5",
"@cite_10"
],
"mid": [
"2048316909",
"62311994",
"2128407051"
],
"abstract": [
"Cumulative citation recommendation refers to the task of filtering a time-ordered corpus for documents that are highly relevant to a predefined set of entities. This task has been introduced at the TREC Knowledge Base Acceleration track in 2012, where two main families of approaches emerged: classification and ranking. In this paper we perform an experimental comparison of these two strategies using supervised learning with a rich feature set. Our main finding is that ranking outperforms classification on all evaluation settings and metrics. Our analysis also reveals that a ranking-based approach has more potential for future improvements.",
"Knowledge bases have become indispensable sources of information. It is therefore critical that they rely on the latest information available and get updated every time new facts surface. Knowledge base acceleration (KBA) systems seek to help humans expand knowledge bases like Wikipedia by automatically recommending edits based on incoming content streams. A core step in this process is that of identifying relevant content, i.e., filtering documents that would imply modifications to the attributes or relations of a given target entity. We propose two multi-step classification approaches for this task that consist of two and three binary classification steps, respectively. Both methods share the same initial component, which is concerned with the identification of entity mentions in documents, while subsequent steps involve identification of documents being relevant and or central to a given entity. Using the evaluation platform of the TREC 2012 KBA track and a rich feature set developed for this particular task, we show that both approaches deliver state-of-the-art performance.",
"Over the past few years, massive amounts of world knowledge have been accumulated in publicly available knowledge bases, such as Freebase, NELL, and YAGO. Yet despite their seemingly huge size, these knowledge bases are greatly incomplete. For example, over 70 of people included in Freebase have no known place of birth, and 99 have no known ethnicity. In this paper, we propose a way to leverage existing Web-search-based question-answering technology to fill in the gaps in knowledge bases in a targeted way. In particular, for each entity attribute, we learn the best set of queries to ask, such that the answer snippets returned by the search engine are most likely to contain the correct value for that attribute. For example, if we want to find Frank Zappa's mother, we could ask the query who is the mother of Frank Zappa'. However, this is likely to return The Mothers of Invention', which was the name of his band. Our system learns that it should (in this case) add disambiguating terms, such as Zappa's place of birth, in order to make it more likely that the search results contain snippets mentioning his mother. Our system also learns how many different queries to ask for each attribute, since in some cases, asking too many can hurt accuracy (by introducing false positives). We discuss how to aggregate candidate answers across multiple queries, ultimately returning probabilistic predictions for possible values for each attribute. Finally, we evaluate our system and show that it is able to extract a large number of facts with high confidence."
]
} |
1703.10344 | 2054434916 | Wikipedia entity pages are a valuable source of information for direct consumption and for knowledge-base construction, update and maintenance. Facts in these entity pages are typically supported by references. Recent studies show that as much as 20 of the references are from online news sources. However, many entity pages are incomplete even if relevant information is already available in existing news articles. Even for the already present references, there is often a delay between the news article publication time and the reference time. In this work, we therefore look at Wikipedia through the lens of news and propose a novel news-article suggestion task to improve news coverage in Wikipedia, and reduce the lag of newsworthy references. Our work finds direct application, as a precursor, to Wikipedia page generation and knowledge-base acceleration tasks that rely on relevant and high quality input sources. We propose a two-stage supervised approach for suggesting news articles to entity pages for a given state of Wikipedia. First, we suggest news articles to Wikipedia entities (article-entity placement) relying on a rich set of features which take into account the salience and relative authority of entities, and the novelty of news articles to entity pages. Second, we determine the exact section in the entity page for the input article (article-section placement) guided by class-based section templates. We perform an extensive evaluation of our approach based on ground-truth data that is extracted from external references in Wikipedia. We achieve a high precision value of up to 93 in the article-entity suggestion stage and upto 84 for the article-section placement. Finally, we compare our approach against competitive baselines and show significant improvements. | Determining which entities are prominent or salient in a given text has a long history in NLP, sparked by the linguistic theory of Centering @cite_19 . Salience has been used in pronoun and co-reference resolution @cite_15 , or to predict which entities will be included in an abstract of an article @cite_4 . Frequent features to measure salience include the frequency of an entity in a document, positioning of an entity, grammatical function or internal entity structure (POS tags, head nouns etc.). These approaches are not currently aimed at knowledge base generation or Wikipedia coverage extension but we postulate that an entity's salience in a news article is a prerequisite to the news article being relevant enough to be included in an entity page. We therefore use the salience features in @cite_4 as part of our model. However, these features are document-internal --- we will show that they are not sufficient to predict news inclusion into an entity page and add features of entity authority, news authority and novelty that measure the relations between several entities, between entity and news article as well as between several competing news articles. | {
"cite_N": [
"@cite_19",
"@cite_15",
"@cite_4"
],
"mid": [
"1489463938",
"2130848543",
"2250818300"
],
"abstract": [
"PART I. GOALS OF CENTERING PART II. UTTERANCE-LEVEL ISSUES IN CENTERING PART III. CENTERING AS A CROSS-LINGUISTIC UNIVERSAL PART IV. THE ROLE OF CENTERING IN PROCESSING MODELS OF DISCOURSE PART V. INFORMATION STRUCTURE AND CENTERING PART VI. DISCOURSE STRUCTURE AND CENTERING",
"The research focus of computational coreference resolution has exhibited a shift from heuristic approaches to machine learning approaches in the past decade. This paper surveys the major milestones in supervised coreference research since its inception fifteen years ago.",
"Although many NLP systems are moving toward entity-based processing, most still identify important phrases using classical keyword-based approaches. To bridge this gap, we introduce the task of entity salience: assigning a relevance score to each entity in a document. We demonstrate how a labeled corpus for the task can be automatically generated from a corpus of documents and accompanying abstracts. We then show how a classifier with features derived from a standard NLP pipeline outperforms a strong baseline by 34 . Finally, we outline initial experiments on further improving accuracy by leveraging background knowledge about the relationships between entities."
]
} |
1703.10304 | 2604550994 | Reconstruction based on the stereo camera has received considerable attention recently, but two particular challenges still remain. The first concerns the need to aggregate similar pixels in an effective approach, and the second is to maintain as much of the available information as possible while ensuring sufficient accuracy. To overcome these issues, we propose a new 3D representation method, namely, planecell, that extracts planarity from the depth-assisted image segmentation and then projects these depth planes into the 3D world. An energy function formulated from Conditional Random Field that generalizes the planar relationships is maximized to merge coplanar segments. We evaluate our method with a variety of reconstruction baselines on both KITTI and Middlebury datasets, and the results indicate the superiorities compared to other 3D space representation methods in accuracy, memory requirements and further applications. | Basic 3D map representation methods using an image pair are inheritors of various stereo matching algorithms @cite_9 @cite_11 @cite_21 @cite_15 @cite_23 @cite_10 . Point-based 3D reconstruction methods directly transforming stereo matching results lack structural representations. Recent point-level online scanning @cite_0 produces a high-quality 3D model of small objects with the geometric surface prior, which is simpler to operate than strong shape assumptions. For large-scale reconstructions, sparse point-based representations are mainly used for their quality and speed. The point-based maps embedded in the system @cite_15 is designed for real-time applications, such as localization. Different features have been developed for this purpose. For example, the ORB feature matching @cite_14 is designed for fast tracking via a binary descriptor. Adopting denser point clouds in the mapping is challenging because it involves managing millions of discrete values. | {
"cite_N": [
"@cite_14",
"@cite_9",
"@cite_21",
"@cite_0",
"@cite_23",
"@cite_15",
"@cite_10",
"@cite_11"
],
"mid": [
"",
"",
"2963502507",
"2075402943",
"2440384215",
"2139900871",
"1932937519",
"2117248802"
],
"abstract": [
"",
"",
"We present a method for extracting depth information from a rectified image pair. Our approach focuses on the first stage of many stereo algorithms: the matching cost computation. We approach the problem by learning a similarity measure on small image patches using a convolutional neural network. Training is carried out in a supervised manner by constructing a binary classification data set with examples of similar and dissimilar pairs of patches. We examine two network architectures for this task: one tuned for speed, the other for accuracy. The output of the convolutional neural network is used to initialize the stereo matching cost. A series of post-processing steps follow: cross-based cost aggregation, semiglobal matching, a left-right consistency check, subpixel enhancement, a median filter, and a bilateral filter. We evaluate our method on the KITTI 2012, KITTI 2015, and Middlebury stereo data sets and show that it outperforms other approaches on all three data sets.",
"We present a combined hardware and software solution for markerless reconstruction of non-rigidly deforming physical objects with arbitrary shape in real-time. Our system uses a single self-contained stereo camera unit built from off-the-shelf components and consumer graphics hardware to generate spatio-temporally coherent 3D models at 30 Hz. A new stereo matching algorithm estimates real-time RGB-D data. We start by scanning a smooth template model of the subject as they move rigidly. This geometric surface prior avoids strong scene assumptions, such as a kinematic human skeleton or a parametric shape model. Next, a novel GPU pipeline performs non-rigid registration of live RGB-D data to the smooth template using an extended non-linear as-rigid-as-possible (ARAP) framework. High-frequency details are fused onto the final mesh using a linear deformation model. The system is an order of magnitude faster than state-of-the-art methods, while matching the quality and robustness of many offline algorithms. We show precise real-time reconstructions of diverse scenes, including: large deformations of users' heads, hands, and upper bodies; fine-scale wrinkles and folds of skin and clothing; and non-rigid interactions performed by users on flexible objects such as toys. We demonstrate how acquired models can be used for many interactive scenarios, including re-texturing, online performance capture and preview, and real-time shape and motion re-targeting.",
"In the past year, convolutional neural networks have been shown to perform extremely well for stereo estimation. However, current architectures rely on siamese networks which exploit concatenation followed by further processing layers, requiring a minute of GPU computation per image pair. In contrast, in this paper we propose a matching network which is able to produce very accurate results in less than a second of GPU computation. Towards this goal, we exploit a product layer which simply computes the inner product between the two representations of a siamese architecture. We train our network by treating the problem as multi-class classification, where the classes are all possible disparities. This allows us to get calibrated scores, which result in much better matching performance when compared to existing approaches.",
"The computation of stereoscopic depth is an important field of computer vision. Although a large variety of algorithms has been developed, the traditional correlation-based versions of these algorithms are prevalent. This is mainly due to easy implementation and handling but also to the linear computational complexity, as compared to more elaborated algorithms based on diffusion processes, graph-cut or bilateral filtering. In this paper, we introduce a new two-stage matching cost for the traditional approach: the summed normalized cross-correlation (SNCC). This new cost function performs a normalized cross-correlation in the first stage and aggregates the correlation values in a second stage. We show that this new measure can be implemented efficiently and that it leads to a substantial improvement of the performance of the traditional stereo approach because it is less sensitive to high contrast outliers.",
"Stereo techniques have witnessed tremendous progress over the last decades, yet some aspects of the problem still remain challenging today. Striking examples are reflecting and textureless surfaces which cannot easily be recovered using traditional local regularizers. In this paper, we therefore propose to regularize over larger distances using object-category specific disparity proposals (displets) which we sample using inverse graphics techniques based on a sparse disparity estimate and a semantic segmentation of the image. The proposed displets encode the fact that objects of certain categories are not arbitrarily shaped but typically exhibit regular structures. We integrate them as non-local regularizer for the challenging object class ‘car’ into a superpixel based CRF framework and demonstrate its benefits on the KITTI stereo evaluation. At time of submission, our approach ranks first across all KITTI stereo leaderboards.",
"This paper describes the semiglobal matching (SGM) stereo method. It uses a pixelwise, mutual information (Ml)-based matching cost for compensating radiometric differences of input images. Pixelwise matching is supported by a smoothness constraint that is usually expressed as a global cost function. SGM performs a fast approximation by pathwise optimizations from all directions. The discussion also addresses occlusion detection, subpixel refinement, and multibaseline matching. Additionally, postprocessing steps for removing outliers, recovering from specific problems of structured environments, and the interpolation of gaps are presented. Finally, strategies for processing almost arbitrarily large images and fusion of disparity images using orthographic projection are proposed. A comparison on standard stereo images shows that SGM is among the currently top-ranked algorithms and is best, if subpixel accuracy is considered. The complexity is linear to the number of pixels and disparity range, which results in a runtime of just 1-2 seconds on typical test images. An in depth evaluation of the Ml-based matching cost demonstrates a tolerance against a wide range of radiometric transformations. Finally, examples of reconstructions from huge aerial frame and pushbroom images demonstrate that the presented ideas are working well on practical problems."
]
} |
1703.10304 | 2604550994 | Reconstruction based on the stereo camera has received considerable attention recently, but two particular challenges still remain. The first concerns the need to aggregate similar pixels in an effective approach, and the second is to maintain as much of the available information as possible while ensuring sufficient accuracy. To overcome these issues, we propose a new 3D representation method, namely, planecell, that extracts planarity from the depth-assisted image segmentation and then projects these depth planes into the 3D world. An energy function formulated from Conditional Random Field that generalizes the planar relationships is maximized to merge coplanar segments. We evaluate our method with a variety of reconstruction baselines on both KITTI and Middlebury datasets, and the results indicate the superiorities compared to other 3D space representation methods in accuracy, memory requirements and further applications. | Recent studies on voxelized 3D reconstruction focus on infusing primitives into the reconstructions @cite_18 @cite_6 @cite_20 @cite_17 or utilizing scalable data structures to meet CPU requirements @cite_13 . Dame al proposed a formulation which combines shape priors-based tracking and reconstruction. The map was represented as voxels with two parameters including the distance to the closest surface and the confidence value. Nonetheless, the accuracy of volumetric reconstruction is always limited to itself, and re-estimating object surfaces from voxels or 3D grids lead to ambiguities. | {
"cite_N": [
"@cite_18",
"@cite_6",
"@cite_13",
"@cite_20",
"@cite_17"
],
"mid": [
"2216747437",
"2462462929",
"2167687475",
"1987737421",
"2091297047"
],
"abstract": [
"In this paper, we propose a structural segmentation algorithm to partition multi-view stereo reconstructed surfaces of large-scale urban environments into structural segments. Each segment corresponds to a structural component describable by a surface primitive of up to the second order. This segmentation is for use in subsequent urban object modeling, vectorization, and recognition. To overcome the high geometrical and topological noise levels in the 3D reconstructed urban surfaces, we formulate the structural segmentation as a higher-order Conditional Random Field (CRF) labeling problem. It not only incorporates classical lower-order 2D and 3D local cues, but also encodes contextual geometric regularities to disambiguate the noisy local cues. A general higher-order CRF is difficult to solve. We develop a bottom-up progressive approach through a patch-based surface representation, which iteratively evolves from the initial mesh triangles to the final segmentation. Each iteration alternates between performing a prior discovery step, which finds the contextual regularities of the patch-based representation, and an inference step that leverages the regularities as higher-order priors to construct a more stable and regular segmentation. The efficiency and robustness of the proposed method is extensively demonstrated on real reconstruction models, yielding significantly better performance than classical mesh segmentation methods.",
"In this paper, we propose a non-local structured prior for volumetric multi-view 3D reconstruction. Towards this goal, we present a novel Markov random field model based on ray potentials in which assumptions about large 3D surface patches such as planarity or Manhattan world constraints can be efficiently encoded as probabilistic priors. We further derive an inference algorithm that reasons jointly about voxels, pixels and image segments, and estimates marginal distributions of appearance, occupancy, depth, normals and planarity. Key to tractable inference is a novel hybrid representation that spans both voxel and pixel space and that integrates non-local information from 2D image segmentations in a principled way. We compare our non-local prior to commonly employed local smoothness assumptions and a variety of state-of-the-art volumetric reconstruction baselines on challenging outdoor scenes with textureless and reflective surfaces. Our experiments indicate that regularizing over larger distances has the potential to resolve ambiguities where local regularizers fail.",
"Our abilities in scene understanding, which allow us to perceive the 3D structure of our surroundings and intuitively recognise the objects we see, are things that we largely take for granted, but for robots, the task of understanding large scenes quickly remains extremely challenging. Recently, scene understanding approaches based on 3D reconstruction and semantic segmentation have become popular, but existing methods either do not scale, fail outdoors, provide only sparse reconstructions or are rather slow. In this paper, we build on a recent hash-based technique for large-scale fusion and an efficient mean-field inference algorithm for densely-connected CRFs to present what to our knowledge is the first system that can perform dense, large-scale, outdoor semantic reconstruction of a scene in (near) real time. We also present a ‘semantic fusion’ approach that allows us to handle dynamic objects more effectively than previous approaches. We demonstrate the effectiveness of our approach on the KITTI dataset, and provide qualitative and quantitative results showing high-quality dense reconstruction and labelling of a number of scenes.",
"A novel method able to recover scene planes of arbitrary position and orientation from oriented images using homographies is presented. Planar regions are reconstructed using only sparse, affine-invariant sets of corresponding seed regions. These regions are iteratively expanded and refined using plane-induced homographies. Experiments on synthetic data show the high accuracy of the reconstruction and demonstrate that the reconstruction method can cope with large baseline changes. Experiments on real images show the performance of the method on practically relevant scenes.",
"In this paper, we present a novel method, the first to date to our knowledge, which is capable of directly and automatically producing a concise and idealized 3D representation from unstructured point data of complex cluttered real-world scenes, with a high level of noise and a significant proportion of outliers, such as those obtained from passive stereo. Our algorithm can digest millions of input points into an optimized lightweight watertight polygonal mesh free of self-intersection, that preserves the structural components of the scene at a user-defined scale, and completes missing scene parts in a plausible manner. To achieve this, our algorithm incorporates priors on urban and architectural scenes, notably the prevalence of vertical structures and orthogonal intersections. A major contribution of our work is an adaptive decomposition of 3D space induced by planar primitives, namely a polyhedral cell complex. We experimentally validate our approach on several challenging noisy point clouds of urban and architectural scenes."
]
} |
1703.10444 | 2604233748 | We consider the problems of robust PAC learning from distributed and streaming data, which may contain malicious errors and outliers, and analyze their fundamental complexity questions. In particular, we establish lower bounds on the communication complexity for distributed robust learning performed on multiple machines, and on the space complexity for robust learning from streaming data on a single machine. These results demonstrate that gaining robustness of learning algorithms is usually at the expense of increased complexities. As far as we know, this work gives the first complexity results for distributed and online robust PAC learning. | Since its introduction in @cite_11 , communication complexity @cite_23 has proven to be a powerful technique for establishing lower bounds in a variety of settings, including distributed @cite_5 @cite_10 @cite_16 and streaming data models @cite_22 . | {
"cite_N": [
"@cite_22",
"@cite_23",
"@cite_5",
"@cite_16",
"@cite_10",
"@cite_11"
],
"mid": [
"1965972569",
"",
"",
"2953101345",
"2162576315",
"2002501531"
],
"abstract": [
"1 Introduction 2 Map 3 The Data Stream Phenomenon 4 Data Streaming: Formal Aspects 5 Foundations: Basic Mathematical Ideas 6 Foundations: Basic Algorithmic Techniques 7 Foundations: Summary 8 Streaming Systems 9 New Directions 10 Historic Notes 11 Concluding Remarks Acknowledgements References.",
"",
"",
"We explore the connection between dimensionality and communication cost in distributed learning problems. Specifically we study the problem of estimating the mean @math of an unknown @math dimensional gaussian distribution in the distributed setting. In this problem, the samples from the unknown distribution are distributed among @math different machines. The goal is to estimate the mean @math at the optimal minimax rate while communicating as few bits as possible. We show that in this setting, the communication cost scales linearly in the number of dimensions i.e. one needs to deal with different dimensions individually. Applying this result to previous lower bounds for one dimension in the interactive setting ZDJW13 and to our improved bounds for the simultaneous setting, we prove new lower bounds of @math and @math for the bits of communication needed to achieve the minimax squared loss, in the interactive and simultaneous settings respectively. To complement, we also demonstrate an interactive protocol achieving the minimax squared loss with @math bits of communication, which improves upon the simple simultaneous protocol by a logarithmic factor. Given the strong lower bounds in the general setting, we initiate the study of the distributed parameter estimation problems with structured parameters. Specifically, when the parameter is promised to be @math -sparse, we show a simple thresholding based protocol that achieves the same squared loss while saving a @math factor of communication. We conjecture that the tradeoff between communication and squared loss demonstrated by this protocol is essentially optimal up to logarithmic factor.",
"We establish lower bounds on minimax risks for distributed statistical estimation under a communication budget. Such lower bounds reveal the minimum amount of communication required by any procedure to achieve the centralized minimax-optimal rates for statistical estimation. We study two classes of protocols: one in which machines send messages independently, and a second allowing for interactive communication. We establish lower bounds for several problems, including various types of location models, as well as for parameter estimation in regression models.",
"Let M e 0, 1, 2, ..., m —1 , N e 0, 1, 2,..., n —1 , and f:M × N → 0, 1 a Boolean-valued function. We will be interested in the following problem and its related questions. Let i e M , j e N be integers known only to two persons P 1 and P 2 , respectively. For P 1 and P 2 to determine cooperatively the value f ( i, j ), they send information to each other alternately, one bit at a time, according to some algorithm. The quantity of interest, which measures the information exchange necessary for computing f , is the minimum number of bits exchanged in any algorithm. For example, if f ( i, j ) e ( i + j ) mod 2. then 1 bit of information (conveying whether i is odd) sent from P 1 to P 2 will enable P 2 to determine f ( i, j ), and this is clearly the best possible. The above problem is a variation of a model of Abelson [1] concerning information transfer in distributive computions."
]
} |
1703.10444 | 2604233748 | We consider the problems of robust PAC learning from distributed and streaming data, which may contain malicious errors and outliers, and analyze their fundamental complexity questions. In particular, we establish lower bounds on the communication complexity for distributed robust learning performed on multiple machines, and on the space complexity for robust learning from streaming data on a single machine. These results demonstrate that gaining robustness of learning algorithms is usually at the expense of increased complexities. As far as we know, this work gives the first complexity results for distributed and online robust PAC learning. | Several recent works on communication complexity for distributed PAC learning include @cite_21 @cite_9 @cite_1 and there are also some works about distributed statistical estimation @cite_5 @cite_10 . In particular, Duchi @cite_7 demonstrated an analysis tool based on the information Fano inequalities. Garg @cite_16 investigated how the communication cost scales in the parameter dimensionality for distributed statistical estimation. Kremer @cite_24 provide a connection between communication complexity and VC-dimension. However, those works generally focus on the case no outliers or malicious errors are presented in the learning process. | {
"cite_N": [
"@cite_7",
"@cite_9",
"@cite_21",
"@cite_1",
"@cite_24",
"@cite_5",
"@cite_16",
"@cite_10"
],
"mid": [
"2106586162",
"",
"2127180992",
"2151390636",
"2089797683",
"",
"2953101345",
"2162576315"
],
"abstract": [
"In this technical note, we give two extensions of the classical Fano inequality in information theory. The first extends Fano's inequality to the setting of estimation, providing lower bounds on the probability that an estimator of a discrete quantity is within some distance @math of the quantity. The second inequality extends our bound to a continuum setting and provides a volume-based bound. We illustrate how these inequalities lead to direct and simple proofs of several statistical minimax lower bounds.",
"",
"We consider the problem of PAC-learning from distributed data and analyze fundamental communication complexity questions involved. We provide general upper and lower bounds on the amount of communication needed to learn well, showing that in addition to VC-dimension and covering number, quantities such as the teaching-dimension and mistake-bound of a class play an important role. We also present tight results for a number of common concept classes including conjunctions, parity functions, and decision lists. For linear separators, we show that for non-concentrated distributions, we can use a version of the Perceptron algorithm to learn with much less communication than the number of updates given by the usual margin bound. We also show how boosting can be performed in a generic manner in the distributed setting to achieve communication with only logarithmic dependence on 1 epsilon for any concept class, and demonstrate how recent work on agnostic learning from class-conditional queries can be used to achieve low communication in agnostic settings as well. We additionally present an analysis of privacy, considering both differential privacy and a notion of distributional privacy that is especially appealing in this context.",
"We consider the problem of learning classifiers for labeled data that has been distributed across several nodes. Our goal is to find a single classifier, with small approximation error, across all datasets while minimizing the communication between nodes. This setting models real-world communication bottlenecks in the processing of massive distributed datasets. We present several very general sampling-based solutions as well as some two-way protocols which have a provable exponential speed-up over any one-way protocol. We focus on core problems for noiseless data distributed across two or more nodes. The techniques we introduce are reminiscent of active learning, but rather than actively probing labels, nodes actively communicate with each other, each node simultaneously learning the important data from another node.",
"",
"",
"We explore the connection between dimensionality and communication cost in distributed learning problems. Specifically we study the problem of estimating the mean @math of an unknown @math dimensional gaussian distribution in the distributed setting. In this problem, the samples from the unknown distribution are distributed among @math different machines. The goal is to estimate the mean @math at the optimal minimax rate while communicating as few bits as possible. We show that in this setting, the communication cost scales linearly in the number of dimensions i.e. one needs to deal with different dimensions individually. Applying this result to previous lower bounds for one dimension in the interactive setting ZDJW13 and to our improved bounds for the simultaneous setting, we prove new lower bounds of @math and @math for the bits of communication needed to achieve the minimax squared loss, in the interactive and simultaneous settings respectively. To complement, we also demonstrate an interactive protocol achieving the minimax squared loss with @math bits of communication, which improves upon the simple simultaneous protocol by a logarithmic factor. Given the strong lower bounds in the general setting, we initiate the study of the distributed parameter estimation problems with structured parameters. Specifically, when the parameter is promised to be @math -sparse, we show a simple thresholding based protocol that achieves the same squared loss while saving a @math factor of communication. We conjecture that the tradeoff between communication and squared loss demonstrated by this protocol is essentially optimal up to logarithmic factor.",
"We establish lower bounds on minimax risks for distributed statistical estimation under a communication budget. Such lower bounds reveal the minimum amount of communication required by any procedure to achieve the centralized minimax-optimal rates for statistical estimation. We study two classes of protocols: one in which machines send messages independently, and a second allowing for interactive communication. We establish lower bounds for several problems, including various types of location models, as well as for parameter estimation in regression models."
]
} |
1703.10444 | 2604233748 | We consider the problems of robust PAC learning from distributed and streaming data, which may contain malicious errors and outliers, and analyze their fundamental complexity questions. In particular, we establish lower bounds on the communication complexity for distributed robust learning performed on multiple machines, and on the space complexity for robust learning from streaming data on a single machine. These results demonstrate that gaining robustness of learning algorithms is usually at the expense of increased complexities. As far as we know, this work gives the first complexity results for distributed and online robust PAC learning. | On the other hand, studies on streaming algorithms focus on the scenario where input arrives very rapidly and there is limited memory to store the input @cite_22 , and investigate the necessary space cost under the most adversarial data orders for various data mining problems, include quantile query @cite_5 , frequent elements query @cite_13 @cite_0 , regression @cite_4 and model monitoring @cite_14 . Those algorithms do not consider the case with malicious errors either. | {
"cite_N": [
"@cite_14",
"@cite_4",
"@cite_22",
"@cite_0",
"@cite_5",
"@cite_13"
],
"mid": [
"",
"2139058364",
"1965972569",
"2064379477",
"",
"2795088498"
],
"abstract": [
"",
"In this paper we give a very space-efficient, yet fast method for estimating the fractal dimensionality of the points in a data stream. Algorithms to estimate the fractal dimension exist, such as the straightforward quadratic algorithm and the faster O(N log N) or even O(N) box-counting algorithms. However, the sub-quadratic algorithms require Ω(N) space. In this paper, we propose an algorithm that computes the fractal dimension in a single pass, using a constant amount of memory relative to data cardinality. Experimental results on synthetic and real world data sets demonstrate the effectiveness of our algorithm.",
"1 Introduction 2 Map 3 The Data Stream Phenomenon 4 Data Streaming: Formal Aspects 5 Foundations: Basic Mathematical Ideas 6 Foundations: Basic Algorithmic Techniques 7 Foundations: Summary 8 Streaming Systems 9 New Directions 10 Historic Notes 11 Concluding Remarks Acknowledgements References.",
"The frequency moments of a sequence containing mi elements of type i, for 1 i n, are the numbers Fk = P n=1 m k . We consider the space complexity of randomized algorithms that approximate the numbers Fk, when the elements of the sequence are given one by one and cannot be stored. Surprisingly, it turns out that the numbers F0;F1 and F2 can be approximated in logarithmic space, whereas the approximation of Fk for k 6 requires n (1) space. Applications to data bases are mentioned as well.",
"",
"We propose an integrated approach for solving both problems of finding the most popular k elements, and finding frequent elements in a data stream. Our technique is efficient and exact if the alphabet under consideration is small. In the more practical large alphabet case, our solution is space efficient and reports both top-k and frequent elements with tight guarantees on errors. For general data distributions, our top-k algorithm can return a set of k' elements, where k' ≃ k, which are guaranteed to be the top-k' elements; and we use minimal space for calculating frequent elements. For realistic Zipfian data, our space requirement for the frequent elements problem decreases dramatically with the parameter of the distribution; and for top-k queries, we ensure that only the top-k elements, in the correct order, are reported. Our experiments show significant space reductions with no loss in accuracy."
]
} |
1703.10584 | 2604670924 | This paper develops and evaluates a new tensor field representation to express the geometric affordance of one object over another. We expand the well known bisector surface representation to one that is weight-driven and that retains the provenance of surface points with directional vectors. We also incorporate the notion of affordance keypoints which allow for faster decisions at a point of query and with a compact and straightforward descriptor. Using a single interaction example, we are able to generalize to previously-unseen scenarios; both synthetic and also real scenes captured with RGBD sensors. We show how our interaction tensor allows for significantly better performance over alternative formulations. Evaluations also include crowdsourcing comparisons that confirm the validity of our affordance proposals, which agree on average 84 of the time with human judgments, and which is 20-40 better than the baseline methods. | An important body of research comes from the developmental robotics field @cite_12 . The core of these approaches is the representation and learning of actions and predicting the consequences of these over a set of objects. These approaches use visual features describing shape, color, size and relative distances to capture object properties and effects. Using robot self-exploration and human demonstrations the systems benefit from single-object affordances to execute more complex interactions and execute a plan (task planning). For instance, @cite_16 shows a robot learning in a self-supervised manner to use a tool by observing the effects of its actions on other objects. | {
"cite_N": [
"@cite_16",
"@cite_12"
],
"mid": [
"1574909006",
"2528967817"
],
"abstract": [
"The ability to learn about and efficiently use tools constitutes a desirable property for general purpose humanoid robots, as it allows them to extend their capabilities beyond the limitations of their own body. Yet, it is a topic that has only recently been tackled from the robotics community. Most of the studies published so far make use of tool representations that allow their models to generalize the knowledge among similar tools in a very limited way. Moreover, most studies assume that the tool is always grasped in its common or canonical grasp position, thus not considering the influence of the grasp configuration in the outcome of the actions performed with them. In the current paper we present a method that tackles both issues simultaneously by using an extended set of functional features and a novel representation of the effect of the tool use. Together, they implicitly account for the grasping configuration and allow the iCub to generalize among tools based on their geometry. Moreover, learning happens in a self-supervised manner: First, the robot autonomously discovers the affordance categories of the tools by clustering the effect of their usage. These categories are subsequently used as a teaching signal to associate visually obtained functional features to the expected tool's affordance. In the experiments, we show how this technique can be effectively used to select, given a tool, the best action to achieve a desired effect.",
"Affordances capture the relationships between a robot and the environment in terms of the actions that the robot is able to perform. The notable characteristic of affordance-based perception is that an object is perceived by what it affords (e.g., graspable and rollable), instead of identities (e.g., name, color, and shape). Affordances play an important role in basic robot capabilities such as recognition, planning, and prediction. The key challenges in affordance research are: 1) how to automatically discover the distinctive features that specify an affordance in an online and incremental manner and 2) how to generalize these features to novel environments. This survey provides an entry point for interested researchers, including: 1) a general overview; 2) classification and critical analysis of existing work; 3) discussion of how affordances are useful in developmental robotics; 4) some open questions about how to use the affordance concept; and 5) a few promising research directions."
]
} |
1703.10584 | 2604670924 | This paper develops and evaluates a new tensor field representation to express the geometric affordance of one object over another. We expand the well known bisector surface representation to one that is weight-driven and that retains the provenance of surface points with directional vectors. We also incorporate the notion of affordance keypoints which allow for faster decisions at a point of query and with a compact and straightforward descriptor. Using a single interaction example, we are able to generalize to previously-unseen scenarios; both synthetic and also real scenes captured with RGBD sensors. We show how our interaction tensor allows for significantly better performance over alternative formulations. Evaluations also include crowdsourcing comparisons that confirm the validity of our affordance proposals, which agree on average 84 of the time with human judgments, and which is 20-40 better than the baseline methods. | Another line of research that has benefited from affordance learning is Human-Robot Interaction @cite_23 @cite_1 @cite_26 @cite_29 @cite_4 @cite_2 @cite_27 . In these studies the main goal is to to perform action recognition in a robot observing humans, usually to predict or anticipate human activities, and in this way assist humans better while they perform everyday tasks. | {
"cite_N": [
"@cite_26",
"@cite_4",
"@cite_29",
"@cite_1",
"@cite_27",
"@cite_23",
"@cite_2"
],
"mid": [
"2074043780",
"210657420",
"",
"2040001722",
"2296494999",
"2009151595",
""
],
"abstract": [
"The ability to learn from human demonstration is essential for robots in human environments. The activity models that the robot builds from observation must take both the human motion and the objects involved into account. Object models designed for this purpose should reflect the role of the object in the activity - its function, or affordances. The main contribution of this paper is to represent object directly in terms of their interaction with human hands, rather than in terms of appearance. This enables the direct representation of object affordances function, while being robust to intra-class differences in appearance. Object hypotheses are first extracted from a video sequence as tracks of associated image segments. The object hypotheses are encoded as strings, where the vocabulary corresponds to different types of interaction with human hands. The similarity between two such object descriptors can be measured using a string kernel. Experiments show these functional descriptors to capture differences and similarities in object affordances function that are not represented by appearance.",
"In order to avoid an expensive manual labelling process or to learn object classes autonomously without human intervention, object discovery techniques have been proposed that extract visually similar objects from weakly labelled videos. However, the problem of discovering small or medium sized objects is largely unexplored. We observe that videos with activities involving human-object interactions can serve as weakly labelled data for such cases. Since neither object appearance nor motion is distinct enough to discover objects in such videos, we propose a framework that samples from a space of algorithms and their parameters to extract sequences of object proposals. Furthermore, we model similarity of objects based on appearance and functionality, which is derived from human and object motion. We show that functionality is an important cue for discovering objects from activities and demonstrate the generality of the model on three challenging RGB-D and RGB datasets.",
"",
"Analyzing affordances has its root in socio-cognitive development of primates. Knowing what the environment, including other agents, can offer in terms of action capabilities is important for our day-to-day interaction and cooperation. In this paper, we will merge two complementary aspects of affordances: from agent-object perspective, what an agent afford to do with an object, and from agent-agent perspective, what an agent can afford to do for other agent, and present a unified notion of Affordance Graph. The graph will encode affordances for a variety of tasks: take, give, pick, put on, put into, show, hide, make accessible, etc. Another novelty will be to incorporate the aspects of effort and perspective-taking in constructing such graph. Hence, the Affordance Graph will tell about the action-capabilities of manipulating the objects among the agents and across the places, along with the information about the required level of efforts and the potential places. We will also demonstrate some interesting applications.",
"For robots, the ability to model human configurations and temporal dynamics is crucial for the task of anticipating future human activities, yet requires conflicting properties: On one hand, we need a detailed high-dimensional description of human configurations to reason about the physical plausibility of the prediction; on the other hand, we need a compact representation to be able to parsimoniously model the relations between the human and the environment. We therefore propose a new model, GP-LCRF, which admits both the high-dimensional and low-dimensional representation of humans. It assumes that the high-dimensional representation is generated from a latent variable corresponding to its lowdimensional representation using a Gaussian process. The generative process not only defines the mapping function between the highand low-dimensional spaces, but also models a distribution of humans embedded as a potential function in GP-LCRF along with other potentials to jointly model the rich context among humans, objects and the activity. Through extensive experiments on activity anticipation, we show that our GP-LCRF consistently outperforms the state-of-the-art results and reduces the predicted human trajectory error by 11.6 .",
"In this paper, we propose a method to recognize human body movements and we combine it with the contextual knowledge of human-robot collaboration scenarios provided by an object affordances framework that associates actions with its effects and the objects involved in them. The aim is to equip humanoid robots with action prediction capabilities, allowing them to anticipate effects as soon as a human partner starts performing a physical action, thus enabling interactions between man and robot to be fast and natural. We consider simple actions that characterize a human-robot collaboration scenario with objects being manipulated on a table: inspired from automatic speech recognition techniques, we train a statistical gesture model in order to recognize those physical gestures in real time. Analogies and differences between the two domains are discussed, highlighting the requirements of an automatic gesture recognizer for robots in order to perform robustly and in real time.",
""
]
} |
1703.10584 | 2604670924 | This paper develops and evaluates a new tensor field representation to express the geometric affordance of one object over another. We expand the well known bisector surface representation to one that is weight-driven and that retains the provenance of surface points with directional vectors. We also incorporate the notion of affordance keypoints which allow for faster decisions at a point of query and with a compact and straightforward descriptor. Using a single interaction example, we are able to generalize to previously-unseen scenarios; both synthetic and also real scenes captured with RGBD sensors. We show how our interaction tensor allows for significantly better performance over alternative formulations. Evaluations also include crowdsourcing comparisons that confirm the validity of our affordance proposals, which agree on average 84 of the time with human judgments, and which is 20-40 better than the baseline methods. | Work has also been done using static imagery, where the affordance or interaction is provided as a label rather than demonstrated. @cite_20 @cite_4 @cite_17 @cite_13 based their work on labeled 2D imagery to predict functional regions or attributes on every day objects. | {
"cite_N": [
"@cite_13",
"@cite_4",
"@cite_20",
"@cite_17"
],
"mid": [
"1900424585",
"210657420",
"1989739075",
"1891689858"
],
"abstract": [
"Affordances are fundamental attributes of objects. Affordances reveal the functionalities of objects and the possible actions that can be performed on them. Understanding affordances is crucial for recognizing human activities in visual data and for robots to interact with the world. In this paper we introduce the new problem of mining the knowledge of semantic affordance: given an object, determining whether an action can be performed on it. This is equivalent to connecting verb nodes and noun nodes in WordNet, or filling an affordance matrix encoding the plausibility of each action-object pair. We introduce a new benchmark with crowdsourced ground truth affordances on 20 PASCAL VOC object classes and 957 action classes. We explore a number of approaches including text mining, visual mining, and collaborative filtering. Our analyses yield a number of significant insights that reveal the most effective ways of collecting knowledge of semantic affordances.",
"In order to avoid an expensive manual labelling process or to learn object classes autonomously without human intervention, object discovery techniques have been proposed that extract visually similar objects from weakly labelled videos. However, the problem of discovering small or medium sized objects is largely unexplored. We observe that videos with activities involving human-object interactions can serve as weakly labelled data for such cases. Since neither object appearance nor motion is distinct enough to discover objects in such videos, we propose a framework that samples from a space of algorithms and their parameters to extract sequences of object proposals. Furthermore, we model similarity of objects based on appearance and functionality, which is derived from human and object motion. We show that functionality is an important cue for discovering objects from activities and demonstrate the generality of the model on three challenging RGB-D and RGB datasets.",
"We revisit the notion of object affordances, an idea that speaks to an object's functional properties more than its class label. We study the problem of spatially localizing affordances in the form of 2D segmentation masks annotated with discrete affordance labels. For example, we use affordance masks to denote on what surfaces a person sits, grabs, and looks at when interacting with a variety of everyday objects (such as chairs, bikes, and TVs). We introduce such a functionally-annotated dataset derived from the PASCAL VOC benchmark and empirically evaluate several approaches for predicting such functionally-relevant object regions. We compare \"blind\" approaches that ignore image data, bottom-up approaches that reason about local surface layout, and top-down approaches that reason about structural constraints between surfaces regions of objects. We show that the difficulty of functional region prediction varies considerably across objects, and that in general, top-down functional object models do well, though there is much room for improvement.",
"Reasoning about objects and their affordances is a fundamental problem for visual intelligence. Most of the previous work casts this problem as a classification task where separate classifiers are trained to label objects, recognize attributes, or assign affordances. In this work, we consider the problem of object affordance reasoning using a knowledge base representation. Diverse information of objects are first harvested from images and other meta-data sources. We then learn a knowledge base (KB) using a Markov Logic Network (MLN). Given the learned KB, we show that a diverse set of visual inference tasks can be done in this unified framework without training separate classifiers, including zero-shot affordance prediction and object recognition given human poses."
]
} |
1703.10355 | 2605073971 | In this paper, we introduce transformations of deep rectifier networks, enabling the conversion of deep rectifier networks into shallow rectifier networks. We subsequently prove that any rectifier net of any depth can be represented by a maximum of a number of functions that can be realized by a shallow network with a single hidden layer. The transformations of both deep rectifier nets and deep residual nets are conducted to demonstrate the advantages of the residual nets over the conventional neural nets and the advantages of the deep neural nets over the shallow neural nets. In summary, for two rectifier nets with different depths but with same total number of hidden units, the corresponding single hidden layer representation of the deeper net is much more complex than the corresponding single hidden representation of the shallower net. Similarly, for a residual net and a conventional rectifier net with the same structure except for the skip connections in the residual net, the corresponding single hidden layer representation of the residual net is much more complex than the corresponding single hidden layer representation of the conventional net. | The superior expressive power of deep residual nets was analysed by @cite_6 who showed that residual nets can be understood as a collection of many paths of various lengths and these paths enable the training of very deep networks by leveraging short paths. Unlike plain neural nets, paths through residual networks vary in length. | {
"cite_N": [
"@cite_6"
],
"mid": [
"2963410064"
],
"abstract": [
"In this work we propose a novel interpretation of residual networks showing that they can be seen as a collection of many paths of differing length. Moreover, residual networks seem to enable very deep networks by leveraging only the short paths during training. To support this observation, we rewrite residual networks as an explicit collection of paths. Unlike traditional models, paths through residual networks vary in length. Further, a lesion study reveals that these paths show ensemble-like behavior in the sense that they do not strongly depend on each other. Finally, and most surprising, most paths are shorter than one might expect, and only the short paths are needed during training, as longer paths do not contribute any gradient. For example, most of the gradient in a residual network with 110 layers comes from paths that are only 10-34 layers deep. Our results reveal one of the key characteristics that seem to enable the training of very deep networks: Residual networks avoid the vanishing gradient problem by introducing short paths which can carry gradient throughout the extent of very deep networks."
]
} |
1703.10545 | 2604413317 | Rating platforms enable large-scale collection of user opinion about items (products, other users, etc.). However, many untrustworthy users give fraudulent ratings for excessive monetary gains. In the paper, we present FairJudge, a system to identify such fraudulent users. We propose three metrics: (i) the fairness of a user that quantifies how trustworthy the user is in rating the products, (ii) the reliability of a rating that measures how reliable the rating is, and (iii) the goodness of a product that measures the quality of the product. Intuitively, a user is fair if it provides reliable ratings that are close to the goodness of the product. We formulate a mutually recursive definition of these metrics, and further address cold start problems and incorporate behavioral properties of users and products in the formulation. We propose an iterative algorithm, FairJudge, to predict the values of the three metrics. We prove that FairJudge is guaranteed to converge in a bounded number of iterations, with linear time complexity. By conducting five different experiments on five rating platforms, we show that FairJudge significantly outperforms nine existing algorithms in predicting fair and unfair users. We reported the 100 most unfair users in the Flipkart network to their review fraud investigators, and 80 users were correctly identified (80 accuracy). The FairJudge algorithm is already being deployed at Flipkart. | Existing works in rating fraud detection can be categorized into network-based and behavior-based algorithms: algorithms are based on iterative learning, belief propagation, and node ranking techniques. Similar to our proposed algorithm, @cite_31 @cite_11 @cite_33 develop iterative algorithms that jointly assign scores in the rating networks based on consensus of ratings - @cite_31 scores each user, review and product, and @cite_11 scores each user and product. FraudEagle @cite_10 is a belief propagation model to rank users, which assumes fraudsters rate good products poorly and bad products positively, and vice-versa for honest users. Random-walk based algorithms have been developed to detect trolls @cite_14 and link farming from collusion on Twitter @cite_0 . @cite_12 @cite_16 @cite_39 identify group of fraudsters based on local neighborhood of the users. A on network-based fraud detection can be found in @cite_35 . | {
"cite_N": [
"@cite_35",
"@cite_14",
"@cite_33",
"@cite_39",
"@cite_0",
"@cite_31",
"@cite_16",
"@cite_10",
"@cite_12",
"@cite_11"
],
"mid": [
"2089554624",
"2415027247",
"2407292501",
"",
"2005556331",
"2112213600",
"",
"1975223096",
"2101447063",
"2101396617"
],
"abstract": [
"Detecting anomalies in data is a vital task, with numerous high-impact applications in areas such as security, finance, health care, and law enforcement. While numerous techniques have been developed in past years for spotting outliers and anomalies in unstructured collections of multi-dimensional points, with graph data becoming ubiquitous, techniques for structured graph data have been of focus recently. As objects in graphs have long-range correlations, a suite of novel technology has been developed for anomaly detection in graph data. This survey aims to provide a general, comprehensive, and structured overview of the state-of-the-art methods for anomaly detection in data represented as graphs. As a key contribution, we give a general framework for the algorithms categorized under various settings: unsupervised versus (semi-)supervised approaches, for static versus dynamic graphs, for attributed versus plain graphs. We highlight the effectiveness, scalability, generality, and robustness aspects of the methods. What is more, we stress the importance of anomaly attribution and highlight the major techniques that facilitate digging out the root cause, or the why', of the detected anomalies for further analysis and sense-making. Finally, we present several real-world applications of graph-based anomaly detection in diverse domains, including financial, auction, computer traffic, and social networks. We conclude our survey with a discussion on open theoretical and practical challenges in the field.",
"Signed social networks have become increasingly important in recent years because of the ability to model trust-based relationships in review sites like Slashdot, Epinions, and Wikipedia. As a result, many traditional network mining problems have been re-visited in the context of networks in which signs are associated with the links. Examples of such problems include community detection, link prediction, and low rank approximation. In this paper, we will examine the problem of ranking nodes in signed networks. In particular, we will design a ranking model, which has a clear physical interpretation in terms of the sign of the edges in the network. Specifically, we propose the Troll-Trust model that models the probability of trustworthiness of individual data sources as an interpretation for the underlying ranking values. We will show the advantages of this approach over a variety of baselines.",
"With the growth of the Internet and E-commerce, bipartite rating networks are ubiquitous. In such bipartite rating networks, there exist two types of entities: the users and the objects, where users give ratings to objects. A fundamental problem in such networks is how to rank the objects by user’s ratings. Although it has been extensively studied in the past decade, the existing algorithms either cannot guarantee convergence, or are not robust to the spammers. In this paper, we propose six new reputation-based algorithms, where the users’ reputation is determined by the aggregated difference between the users’ ratings and the corresponding objects’ rankings. We prove that all of our algorithms converge into a unique fixed point. The time and space complexity of our algorithms are linear w.r.t. the size of the graph, thus they can be scalable to large datasets. Moreover, our algorithms are robust to the spamming users. We evaluate our algorithms using three real datasets. The experimental results confirm the effectiveness, efficiency, and robustness of our algorithms.",
"",
"Recently, Twitter has emerged as a popular platform for discovering real-time information on the Web, such as news stories and people's reaction to them. Like the Web, Twitter has become a target for link farming, where users, especially spammers, try to acquire large numbers of follower links in the social network. Acquiring followers not only increases the size of a user's direct audience, but also contributes to the perceived influence of the user, which in turn impacts the ranking of the user's tweets by search engines. In this paper, we first investigate link farming in the Twitter network and then explore mechanisms to discourage the activity. To this end, we conducted a detailed analysis of links acquired by over 40,000 spammer accounts suspended by Twitter. We find that link farming is wide spread and that a majority of spammers' links are farmed from a small fraction of Twitter users, the social capitalists, who are themselves seeking to amass social capital and links by following back anyone who follows them. Our findings shed light on the social dynamics that are at the root of the link farming problem in Twitter network and they have important implications for future designs of link spam defenses. In particular, we show that a simple user ranking scheme that penalizes users for connecting to spammers can effectively address the problem by disincentivizing users from linking with other users simply to gain influence.",
"Online reviews provide valuable information about products and services to consumers. However, spammers are joining the community trying to mislead readers by writing fake reviews. Previous attempts for spammer detection used reviewers' behaviors, text similarity, linguistics features and rating patterns. Those studies are able to identify certain types of spammers, e.g., those who post many similar reviews about one target entity. However, in reality, there are other kinds of spammers who can manipulate their behaviors to act just like genuine reviewers, and thus cannot be detected by the available techniques. In this paper, we propose a novel concept of a heterogeneous review graph to capture the relationships among reviewers, reviews and stores that the reviewers have reviewed. We explore how interactions between nodes in this graph can reveal the cause of spam and propose an iterative model to identify suspicious reviewers. This is the first time such intricate relationships have been identified for review spam detection. We also develop an effective computation method to quantify the trustiness of reviewers, the honesty of reviews, and the reliability of stores. Different from existing approaches, we don't use review text information. Our model is thus complementary to existing approaches and able to find more difficult and subtle spamming activities, which are agreed upon by human judges after they evaluate our results.",
"",
"Online reviews are an important source for consumers to evaluate products services on the Internet (e.g. Amazon, Yelp, etc.). However, more and more fraudulent reviewers write fake reviews to mislead users. To maximize their impact and share effort, many spam attacks are organized as campaigns, by a group of spammers. In this paper, we propose a new two-step method to discover spammer groups and their targeted products. First, we introduce NFS (Network Footprint Score), a new measure that quantifies the likelihood of products being spam campaign targets. Second, we carefully devise GroupStrainer to cluster spammers on a 2-hop subgraph induced by top ranking products. Our approach has four key advantages: (i) unsupervised detection; both steps require no labeled data, (ii) adversarial robustness; we quantify statistical distortions in the review network, of which spammers have only a partial view, and avoid any side information that spammers can easily evade, (iii) sensemaking; the output facilitates the exploration of the nested hierarchy (i.e., organization) among the spammers, and finally (iv) scalability; both steps have complexity linear in network size, moreover, GroupStrainer operates on a carefully induced subnetwork. We demonstrate the efficiency and effectiveness of our approach on both synthetic and real-world datasets from two different domains with millions of products and reviewers. Moreover, we discover interesting strategies that spammers employ through case studies of our detected groups.",
"Given a directed graph of millions of nodes, how can we automatically spot anomalous, suspicious nodes, judging only from their connectivity patterns? Suspicious graph patterns show up in many applications, from Twitter users who buy fake followers, manipulating the social network, to botnet members performing distributed denial of service attacks, disturbing the network traffic graph. We propose a fast and effective method, CatchSync, which exploits two of the tell-tale signs left in graphs by fraudsters: (a) synchronized behavior: suspicious nodes have extremely similar behavior pattern, because they are often required to perform some task together (such as follow the same user); and (b) rare behavior: their connectivity patterns are very different from the majority. We introduce novel measures to quantify both concepts (\"synchronicity\" and \"normality\") and we propose a parameter-free algorithm that works on the resulting synchronicity-normality plots. Thanks to careful design, CatchSync has the following desirable properties: (a) it is scalable to large datasets, being linear on the graph size; (b) it is parameter free; and (c) it is side-information-oblivious: it can operate using only the topology, without needing labeled data, nor timing information, etc., while still capable of using side information, if available. We applied CatchSync on two large, real datasets 1-billion-edge Twitter social graph and 3-billion-edge Tencent Weibo social graph, and several synthetic ones; CatchSync consistently outperforms existing competitors, both in detection accuracy by 36 on Twitter and 20 on Tencent Weibo, as well as in speed.",
"Many real-life graphs such as social networks and peer-to-peer networks capture the relationships among the nodes by using trust scores to label the edges. Important usage of such networks includes trust prediction, finding the most reliable or trusted node in a local subgraph, etc. For many of these applications, it is crucial to assess the prestige and bias of a node. The bias of a node denotes its propensity to trust mistrust its neighbours and is closely related to truthfulness. If a node trusts all its neighbours, its recommendation of another node as trustworthy is less reliable. It is based on the idea that the recommendation of a highly biased node should weigh less. In this paper, we propose an algorithm to compute the bias and prestige of nodes in networks where the edge weight denotes the trust score. Unlike most other graph-based algorithms, our method works even when the edge weights are not necessarily positive. The algorithm is iterative and runs in O(km) time where k is the number of iterations and m is the total number of edges in the network. The algorithm exhibits several other desirable properties. It converges to a unique value very quickly. Also, the error in bias and prestige values at any particular iteration is bounded. Further, experiments show that our model conforms well to social theories such as the balance theory (enemy of a friend is an enemy, etc.)."
]
} |
1703.10545 | 2604413317 | Rating platforms enable large-scale collection of user opinion about items (products, other users, etc.). However, many untrustworthy users give fraudulent ratings for excessive monetary gains. In the paper, we present FairJudge, a system to identify such fraudulent users. We propose three metrics: (i) the fairness of a user that quantifies how trustworthy the user is in rating the products, (ii) the reliability of a rating that measures how reliable the rating is, and (iii) the goodness of a product that measures the quality of the product. Intuitively, a user is fair if it provides reliable ratings that are close to the goodness of the product. We formulate a mutually recursive definition of these metrics, and further address cold start problems and incorporate behavioral properties of users and products in the formulation. We propose an iterative algorithm, FairJudge, to predict the values of the three metrics. We prove that FairJudge is guaranteed to converge in a bounded number of iterations, with linear time complexity. By conducting five different experiments on five rating platforms, we show that FairJudge significantly outperforms nine existing algorithms in predicting fair and unfair users. We reported the 100 most unfair users in the Flipkart network to their review fraud investigators, and 80 users were correctly identified (80 accuracy). The FairJudge algorithm is already being deployed at Flipkart. | algorithms are often feature-based. Consensus based features have been proposed in @cite_29 @cite_15 -- our proposed goodness metric is also inspired by consensus or wisdom of crowds'. Commonly used features are derived from timestamps @cite_1 @cite_6 @cite_25 and review text @cite_36 @cite_8 @cite_13 . SpEagle @cite_34 extends FraudEagle @cite_10 to incorporate behavior features. BIRDNEST @cite_23 creates a Bayesian model to estimate the belief of each user's deviation in rating behavior from global expected behavior. @cite_12 @cite_26 @cite_27 study coordinated spam behavior of multiple users. A on behavior based algorithms can be found in @cite_9 . | {
"cite_N": [
"@cite_26",
"@cite_8",
"@cite_36",
"@cite_10",
"@cite_29",
"@cite_9",
"@cite_1",
"@cite_6",
"@cite_27",
"@cite_23",
"@cite_15",
"@cite_34",
"@cite_13",
"@cite_25",
"@cite_12"
],
"mid": [
"",
"",
"2121290223",
"1975223096",
"",
"2286850974",
"2136710010",
"",
"",
"2174402890",
"",
"2064058256",
"",
"",
"2101447063"
],
"abstract": [
"",
"",
"Online reviews have been popularly adopted in many applications. Since they can either promote or harm the reputation of a product or a service, buying and selling fake reviews becomes a profitable business and a big threat. In this paper, we introduce a very simple, but powerful review spamming technique that could fail the existing feature-based detection algorithms easily. It uses one truthful review as a template, and replaces its sentences with those from other reviews in a repository. Fake reviews generated by this mechanism are extremely hard to detect: Both the state-of-the-art computational approaches and human readers acquire an error rate of 35 -48 , just slightly better than a random guess. While it is challenging to detect such fake reviews, we have made solid progress in suppressing them. A novel defense method that leverages the difference of semantic flows between synthetic and truthful reviews is developed, which is able to reduce the detection error rate to approximately 22 , a significant improvement over the performance of existing approaches. Nevertheless, it is still a challenging research task to further decrease the error rate. Synthetic Review Spamming Demo: www.cs.ucsb.edu alex_morales reviewspam",
"Online reviews are an important source for consumers to evaluate products services on the Internet (e.g. Amazon, Yelp, etc.). However, more and more fraudulent reviewers write fake reviews to mislead users. To maximize their impact and share effort, many spam attacks are organized as campaigns, by a group of spammers. In this paper, we propose a new two-step method to discover spammer groups and their targeted products. First, we introduce NFS (Network Footprint Score), a new measure that quantifies the likelihood of products being spam campaign targets. Second, we carefully devise GroupStrainer to cluster spammers on a 2-hop subgraph induced by top ranking products. Our approach has four key advantages: (i) unsupervised detection; both steps require no labeled data, (ii) adversarial robustness; we quantify statistical distortions in the review network, of which spammers have only a partial view, and avoid any side information that spammers can easily evade, (iii) sensemaking; the output facilitates the exploration of the nested hierarchy (i.e., organization) among the spammers, and finally (iv) scalability; both steps have complexity linear in network size, moreover, GroupStrainer operates on a carefully induced subnetwork. We demonstrate the efficiency and effectiveness of our approach on both synthetic and real-world datasets from two different domains with millions of products and reviewers. Moreover, we discover interesting strategies that spammers employ through case studies of our detected groups.",
"",
"Different real-world applications have varying definitions of suspicious behaviors. Detection methods often look for the most suspicious parts of the data by optimizing scores, but quantifying the suspiciousness of a behavioral pattern is still an open issue.",
"Online reviews play a crucial role in today's electronic commerce. It is desirable for a customer to read reviews of products or stores before making the decision of what or from where to buy. Due to the pervasive spam reviews, customers can be misled to buy low-quality products, while decent stores can be defamed by malicious reviews. We observe that, in reality, a great portion (> 90 in the data we study) of the reviewers write only one review (singleton review). These reviews are so enormous in number that they can almost determine a store's rating and impression. However, existing methods did not examine this larger part of the reviews. Are most of these singleton reviews truthful ones? If not, how to detect spam reviews in singleton reviews? We call this problem singleton review spam detection. To address this problem, we observe that the normal reviewers' arrival pattern is stable and uncorrelated to their rating pattern temporally. In contrast, spam attacks are usually bursty and either positively or negatively correlated to the rating. Thus, we propose to detect such attacks via unusually correlated temporal patterns. We identify and construct multidimensional time series based on aggregate statistics, in order to depict and mine such correlations. In this way, the singleton review spam detection problem is mapped to a abnormally correlated pattern detection problem. We propose a hierarchical algorithm to robustly detect the time windows where such attacks are likely to have happened. The algorithm also pinpoints such windows in different time resolutions to facilitate faster human inspection. Experimental results show that the proposed method is effective in detecting singleton review attacks. We discover that singleton review is a significant source of spam reviews and largely affects the ratings of online stores.",
"",
"",
"Review fraud is a pervasive problem in online commerce, in which fraudulent sellers write or purchase fake reviews to manipulate perception of their products and services. Fake reviews are often detected based on several signs, including 1) they occur in short bursts of time; 2) fraudulent user accounts have skewed rating distributions. However, these may both be true in any given dataset. Hence, in this paper, we propose an approach for detecting fraudulent reviews which combines these 2 approaches in a principled manner, allowing successful detection even when one of these signs is not present. To combine these 2 approaches, we formulate our Bayesian Inference for Rating Data (BIRD) model, a flexible Bayesian model of user rating behavior. Based on our model we formulate a likelihood-based suspiciousness metric, Normalized Expected Surprise Total (NEST). We propose a linear-time algorithm for performing Bayesian inference using our model and computing the metric. Experiments on real data show that BIRDNEST successfully spots review fraud in large, real-world graphs: the 50 most suspicious users of the Flipkart platform flagged by our algorithm were investigated and all identified as fraudulent by domain experts at Flipkart.",
"",
"Online reviews capture the testimonials of \"real\" people and help shape the decisions of other consumers. Due to the financial gains associated with positive reviews, however, opinion spam has become a widespread problem, with often paid spam reviewers writing fake reviews to unjustly promote or demote certain products or businesses. Existing approaches to opinion spam have successfully but separately utilized linguistic clues of deception, behavioral footprints, or relational ties between agents in a review system. In this work, we propose a new holistic approach called SPEAGLE that utilizes clues from all metadata (text, timestamp, rating) as well as relational data (network), and harness them collectively under a unified framework to spot suspicious users and reviews, as well as products targeted by spam. Moreover, our method can efficiently and seamlessly integrate semi-supervision, i.e., a (small) set of labels if available, without requiring any training or changes in its underlying algorithm. We demonstrate the effectiveness and scalability of SPEAGLE on three real-world review datasets from Yelp.com with filtered (spam) and recommended (non-spam) reviews, where it significantly outperforms several baselines and state-of-the-art methods. To the best of our knowledge, this is the largest scale quantitative evaluation performed to date for the opinion spam problem.",
"",
"",
"Given a directed graph of millions of nodes, how can we automatically spot anomalous, suspicious nodes, judging only from their connectivity patterns? Suspicious graph patterns show up in many applications, from Twitter users who buy fake followers, manipulating the social network, to botnet members performing distributed denial of service attacks, disturbing the network traffic graph. We propose a fast and effective method, CatchSync, which exploits two of the tell-tale signs left in graphs by fraudsters: (a) synchronized behavior: suspicious nodes have extremely similar behavior pattern, because they are often required to perform some task together (such as follow the same user); and (b) rare behavior: their connectivity patterns are very different from the majority. We introduce novel measures to quantify both concepts (\"synchronicity\" and \"normality\") and we propose a parameter-free algorithm that works on the resulting synchronicity-normality plots. Thanks to careful design, CatchSync has the following desirable properties: (a) it is scalable to large datasets, being linear on the graph size; (b) it is parameter free; and (c) it is side-information-oblivious: it can operate using only the topology, without needing labeled data, nor timing information, etc., while still capable of using side information, if available. We applied CatchSync on two large, real datasets 1-billion-edge Twitter social graph and 3-billion-edge Tencent Weibo social graph, and several synthetic ones; CatchSync consistently outperforms existing competitors, both in detection accuracy by 36 on Twitter and 20 on Tencent Weibo, as well as in speed."
]
} |
1703.10553 | 2951675964 | Lossy image compression is generally formulated as a joint rate-distortion optimization to learn encoder, quantizer, and decoder. However, the quantizer is non-differentiable, and discrete entropy estimation usually is required for rate control. These make it very challenging to develop a convolutional network (CNN)-based image compression system. In this paper, motivated by that the local information content is spatially variant in an image, we suggest that the bit rate of the different parts of the image should be adapted to local content. And the content aware bit rate is allocated under the guidance of a content-weighted importance map. Thus, the sum of the importance map can serve as a continuous alternative of discrete entropy estimation to control compression rate. And binarizer is adopted to quantize the output of encoder due to the binarization scheme is also directly defined by the importance map. Furthermore, a proxy function is introduced for binary operation in backward propagation to make it differentiable. Therefore, the encoder, decoder, binarizer and importance map can be jointly optimized in an end-to-end manner by using a subset of the ImageNet database. In low bit rate image compression, experiments show that our system significantly outperforms JPEG and JPEG 2000 by structural similarity (SSIM) index, and can produce the much better visual result with sharp edges, rich textures, and fewer artifacts. | For the existing image standards, , JPEG and JPEG 2000, the codecs actually are separately optimized. In the encoding stage, they first perform a linear transform to an image. Quantization and lossless entropy coding are then utilized to minimize the compression rate. For example, JPEG @cite_21 applies discrete cosine transform (DCT) on @math image patches, quantizes the frequency components and compresses the quantized codes with a variant of Huffman encoding. JPEG 2000 @cite_2 uses a multi-scale orthogonal wavelet decomposition to transform an image, and encodes the quantized codes with the Embedded Block Coding with Optimal Truncation. In the decoding stage, decoding algorithm and inverse transform are designed to minimize distortion. In contrast, we model image compression as a joint rate-distortion optimization, where both nonlinear encoder and decoder are jointly trained in an end-to-end manner. | {
"cite_N": [
"@cite_21",
"@cite_2"
],
"mid": [
"2140196014",
"2099563019"
],
"abstract": [
"A joint ISO CCITT committee known as JPEG (Joint Photographic Experts Group) has been working to establish the first international compression standard for continuous-tone still images, both grayscale and color. JPEG's proposed standard aims to be generic, to support a wide variety of applications for continuous-tone images. To meet the differing needs of many applications, the JPEG standard includes two basic compression methods, each with various modes of operation. A DCT (discrete cosine transform)-based method is specified for 'lossy' compression, and a predictive method for 'lossless' compression. JPEG features a simple lossy technique known as the Baseline method, a subset of the other DCT-based modes of operation. The Baseline method has been by far the most widely implemented JPEG method to date, and is sufficient in its own right for a large number of applications. The author provides an overview of the JPEG standard, and focuses in detail on the Baseline method. >",
"One of the aims of the standardization committee has been the development of Part I, which could be used on a royalty- and fee-free basis. This is important for the standard to become widely accepted. The standardization process, which is coordinated by the JTCI SC29 WG1 of the ISO IEC has already produced the international standard (IS) for Part I. In this article the structure of Part I of the JPFG 2000 standard is presented and performance comparisons with established standards are reported. This article is intended to serve as a tutorial for the JPEG 2000 standard. The main application areas and their requirements are given. The architecture of the standard follows with the description of the tiling, multicomponent transformations, wavelet transforms, quantization and entropy coding. Some of the most significant features of the standard are presented, such as region-of-interest coding, scalability, visual weighting, error resilience and file format aspects. Finally, some comparative results are reported and the future parts of the standard are discussed."
]
} |
1703.10553 | 2951675964 | Lossy image compression is generally formulated as a joint rate-distortion optimization to learn encoder, quantizer, and decoder. However, the quantizer is non-differentiable, and discrete entropy estimation usually is required for rate control. These make it very challenging to develop a convolutional network (CNN)-based image compression system. In this paper, motivated by that the local information content is spatially variant in an image, we suggest that the bit rate of the different parts of the image should be adapted to local content. And the content aware bit rate is allocated under the guidance of a content-weighted importance map. Thus, the sum of the importance map can serve as a continuous alternative of discrete entropy estimation to control compression rate. And binarizer is adopted to quantize the output of encoder due to the binarization scheme is also directly defined by the importance map. Furthermore, a proxy function is introduced for binary operation in backward propagation to make it differentiable. Therefore, the encoder, decoder, binarizer and importance map can be jointly optimized in an end-to-end manner by using a subset of the ImageNet database. In low bit rate image compression, experiments show that our system significantly outperforms JPEG and JPEG 2000 by structural similarity (SSIM) index, and can produce the much better visual result with sharp edges, rich textures, and fewer artifacts. | Recently, several deep learning based image compression models have been developed. For lossless image compression, deep learning models have achieved state-of-the-art performance @cite_17 @cite_1 . For lossy image compression, Toderici al @cite_22 present a recurrent neural network (RNN) to compress @math images. Toderici al @cite_15 further introduce a set of full-resolution compression methods for progressive encoding and decoding of images. These methods learn the compression models by minimizing the distortion for a given compression rate. While our model is end-to-end trained via joint rate-distortion optimization. | {
"cite_N": [
"@cite_15",
"@cite_1",
"@cite_22",
"@cite_17"
],
"mid": [
"2516038988",
"2953318193",
"2276024283",
"2953250761"
],
"abstract": [
"This paper presents a set of full-resolution lossy image compression methods based on neural networks. Each of the architectures we describe can provide variable compression rates during deployment without requiring retraining of the network: each network need only be trained once. All of our architectures consist of a recurrent neural network (RNN)-based encoder and decoder, a binarizer, and a neural network for entropy coding. We compare RNN types (LSTM, associative LSTM) and introduce a new hybrid of GRU and ResNet. We also study \"one-shot\" versus additive reconstruction architectures and introduce a new scaled-additive framework. We compare to previous work, showing improvements of 4.3 -8.8 AUC (area under the rate-distortion curve), depending on the perceptual metric used. As far as we know, this is the first neural network architecture that is able to outperform JPEG at image compression across most bitrates on the rate-distortion curve on the Kodak dataset images, with and without the aid of entropy coding.",
"Modeling the distribution of natural images is a landmark problem in unsupervised learning. This task requires an image model that is at once expressive, tractable and scalable. We present a deep neural network that sequentially predicts the pixels in an image along the two spatial dimensions. Our method models the discrete probability of the raw pixel values and encodes the complete set of dependencies in the image. Architectural novelties include fast two-dimensional recurrent layers and an effective use of residual connections in deep recurrent networks. We achieve log-likelihood scores on natural images that are considerably better than the previous state of the art. Our main results also provide benchmarks on the diverse ImageNet dataset. Samples generated from the model appear crisp, varied and globally coherent.",
"A large fraction of Internet traffic is now driven by requests from mobile devices with relatively small screens and often stringent bandwidth requirements. Due to these factors, it has become the norm for modern graphics-heavy websites to transmit low-resolution, low-bytecount image previews (thumbnails) as part of the initial page load process to improve apparent page responsiveness. Increasing thumbnail compression beyond the capabilities of existing codecs is therefore a current research focus, as any byte savings will significantly enhance the experience of mobile device users. Toward this end, we propose a general framework for variable-rate image compression and a novel architecture based on convolutional and deconvolutional LSTM recurrent networks. Our models address the main issues that have prevented autoencoder neural networks from competing with existing image compression algorithms: (1) our networks only need to be trained once (not per-image), regardless of input image dimensions and the desired compression rate; (2) our networks are progressive, meaning that the more bits are sent, the more accurate the image reconstruction; and (3) the proposed architecture is at least as efficient as a standard purpose-trained autoencoder for a given number of bits. On a large-scale benchmark of 32 @math 32 thumbnails, our LSTM-based approaches provide better visual quality than (headerless) JPEG, JPEG2000 and WebP, with a storage size that is reduced by 10 or more.",
"Modeling the distribution of natural images is challenging, partly because of strong statistical dependencies which can extend over hundreds of pixels. Recurrent neural networks have been successful in capturing long-range dependencies in a number of problems but only recently have found their way into generative image models. We here introduce a recurrent image model based on multi-dimensional long short-term memory units which are particularly suited for image modeling due to their spatial structure. Our model scales to images of arbitrary size and its likelihood is computationally tractable. We find that it outperforms the state of the art in quantitative comparisons on several image datasets and produces promising results when used for texture synthesis and inpainting."
]
} |
1703.10553 | 2951675964 | Lossy image compression is generally formulated as a joint rate-distortion optimization to learn encoder, quantizer, and decoder. However, the quantizer is non-differentiable, and discrete entropy estimation usually is required for rate control. These make it very challenging to develop a convolutional network (CNN)-based image compression system. In this paper, motivated by that the local information content is spatially variant in an image, we suggest that the bit rate of the different parts of the image should be adapted to local content. And the content aware bit rate is allocated under the guidance of a content-weighted importance map. Thus, the sum of the importance map can serve as a continuous alternative of discrete entropy estimation to control compression rate. And binarizer is adopted to quantize the output of encoder due to the binarization scheme is also directly defined by the importance map. Furthermore, a proxy function is introduced for binary operation in backward propagation to make it differentiable. Therefore, the encoder, decoder, binarizer and importance map can be jointly optimized in an end-to-end manner by using a subset of the ImageNet database. In low bit rate image compression, experiments show that our system significantly outperforms JPEG and JPEG 2000 by structural similarity (SSIM) index, and can produce the much better visual result with sharp edges, rich textures, and fewer artifacts. | The most related work is that of @cite_11 @cite_23 based on convolutional autoencoders. Ball 'e al @cite_11 use generalized divisive normalization (GDN) for joint nonlinearity, and replace rounding quantization with additive uniform noise for continuous relaxation of distortion and entropy rate loss. Theis al @cite_23 adopt a smooth approximation of the derivative of the rounding function, and upper-bound the discrete entropy rate loss for continuous relaxation. Our content-weighted image compression system is different with @cite_11 @cite_23 in rate loss, quantization, and continuous relaxation. Instead of rounding and entropy, we define our rate loss on importance map and adopt a simple binarizer for quantization. Moreover, the code length after quantization is spatially invariant in @cite_11 @cite_23 . In contrast, the local code length in our model is content-aware, which is useful in improving visual quality. | {
"cite_N": [
"@cite_23",
"@cite_11"
],
"mid": [
"2950237263",
"2552465432"
],
"abstract": [
"We propose a new approach to the problem of optimizing autoencoders for lossy image compression. New media formats, changing hardware technology, as well as diverse requirements and content types create a need for compression algorithms which are more flexible than existing codecs. Autoencoders have the potential to address this need, but are difficult to optimize directly due to the inherent non-differentiabilty of the compression loss. We here show that minimal changes to the loss are sufficient to train deep autoencoders competitive with JPEG 2000 and outperforming recently proposed approaches based on RNNs. Our network is furthermore computationally efficient thanks to a sub-pixel architecture, which makes it suitable for high-resolution images. This is in contrast to previous work on autoencoders for compression using coarser approximations, shallower architectures, computationally expensive methods, or focusing on small images.",
"We describe an image compression method, consisting of a nonlinear analysis transformation, a uniform quantizer, and a nonlinear synthesis transformation. The transforms are constructed in three successive stages of convolutional linear filters and nonlinear activation functions. Unlike most convolutional neural networks, the joint nonlinearity is chosen to implement a form of local gain control, inspired by those used to model biological neurons. Using a variant of stochastic gradient descent, we jointly optimize the entire model for rate-distortion performance over a database of training images, introducing a continuous proxy for the discontinuous loss function arising from the quantizer. Under certain conditions, the relaxed loss function may be interpreted as the log likelihood of a generative model, as implemented by a variational autoencoder. Unlike these models, however, the compression model must operate at any given point along the rate-distortion curve, as specified by a trade-off parameter. Across an independent set of test images, we find that the optimized method generally exhibits better rate-distortion performance than the standard JPEG and JPEG 2000 compression methods. More importantly, we observe a dramatic improvement in visual quality for all images at all bit rates, which is supported by objective quality estimates using MS-SSIM."
]
} |
1703.10553 | 2951675964 | Lossy image compression is generally formulated as a joint rate-distortion optimization to learn encoder, quantizer, and decoder. However, the quantizer is non-differentiable, and discrete entropy estimation usually is required for rate control. These make it very challenging to develop a convolutional network (CNN)-based image compression system. In this paper, motivated by that the local information content is spatially variant in an image, we suggest that the bit rate of the different parts of the image should be adapted to local content. And the content aware bit rate is allocated under the guidance of a content-weighted importance map. Thus, the sum of the importance map can serve as a continuous alternative of discrete entropy estimation to control compression rate. And binarizer is adopted to quantize the output of encoder due to the binarization scheme is also directly defined by the importance map. Furthermore, a proxy function is introduced for binary operation in backward propagation to make it differentiable. Therefore, the encoder, decoder, binarizer and importance map can be jointly optimized in an end-to-end manner by using a subset of the ImageNet database. In low bit rate image compression, experiments show that our system significantly outperforms JPEG and JPEG 2000 by structural similarity (SSIM) index, and can produce the much better visual result with sharp edges, rich textures, and fewer artifacts. | Our work is also related to binarized neural network (BNN) @cite_3 , where both weights and activations are binarized to @math or @math to save memory storage and run time. Courbariaux al @cite_3 adopt a straight-through estimator to compute the gradient of the binarizer. In our compression system, only the encoder output is binarized to @math or @math , and a similar proxy function is used in backward propagation. | {
"cite_N": [
"@cite_3"
],
"mid": [
"2319920447"
],
"abstract": [
"We introduce a method to train Binarized Neural Networks (BNNs) - neural networks with binary weights and activations at run-time. At training-time the binary weights and activations are used for computing the parameters gradients. During the forward pass, BNNs drastically reduce memory size and accesses, and replace most arithmetic operations with bit-wise operations, which is expected to substantially improve power-efficiency. To validate the effectiveness of BNNs we conduct two sets of experiments on the Torch7 and Theano frameworks. On both, BNNs achieved nearly state-of-the-art results over the MNIST, CIFAR-10 and SVHN datasets. Last but not least, we wrote a binary matrix multiplication GPU kernel with which it is possible to run our MNIST BNN 7 times faster than with an unoptimized GPU kernel, without suffering any loss in classification accuracy. The code for training and running our BNNs is available on-line."
]
} |
1703.10664 | 2604856029 | Deep learning has been demonstrated to achieve excellent results for image classification and object detection. However, the impact of deep learning on video analysis (e.g. action detection and recognition) has been limited due to complexity of video data and lack of annotations. Previous convolutional neural networks (CNN) based video action detection approaches usually consist of two major steps: frame-level action proposal detection and association of proposals across frames. Also, these methods employ two-stream CNN framework to handle spatial and temporal feature separately. In this paper, we propose an end-to-end deep network called Tube Convolutional Neural Network (T-CNN) for action detection in videos. The proposed architecture is a unified network that is able to recognize and localize action based on 3D convolution features. A video is first divided into equal length clips and for each clip a set of tube proposals are generated next based on 3D Convolutional Network (ConvNet) features. Finally, the tube proposals of different clips are linked together employing network flow and spatio-temporal action detection is performed using these linked video proposals. Extensive experiments on several video datasets demonstrate the superior performance of T-CNN for classifying and localizing actions in both trimmed and untrimmed videos compared to state-of-the-arts. | Compared to action recognition, action detection is a more challenging problem @cite_6 @cite_14 , which has been an active area of research. Ke al @cite_1 present an approach for event detection in crowded videos. Tian al @cite_21 develop Spatio-temporal Deformable Parts Model @cite_22 to detect actions in videos. Jain al @cite_27 and Soomro al @cite_5 use supervoxel and selective search to localize the action boundaries. Recently, researchers have leveraged the power of deep learning for action detection. Authors in @cite_18 extract frame-level action proposals using selective search and link them using Viterbi algorithm. While in @cite_11 frame-level action proposals are obtained by EdgeBox and linked by a tracking algorithm. Two-stream R-CNNs for action detection is proposed in @cite_0 , where a spatial Region Proposal Network (RPN) and a motion RPN are used to generate frame-level action proposals. However, these deep learning based approaches detect actions by linking frame-level action proposals and treat the spatial and temporal features of a video separately by training two-stream CNN. Therefore, the temporal consistency in videos is not well explored in the network. In contrast, we determine action tube proposals directly from input videos and extract compact and more effective spatio-temporal features using 3D CNN. | {
"cite_N": [
"@cite_18",
"@cite_14",
"@cite_22",
"@cite_21",
"@cite_1",
"@cite_6",
"@cite_0",
"@cite_27",
"@cite_5",
"@cite_11"
],
"mid": [
"1923332106",
"2095242101",
"2120419212",
"2095661305",
"2137981002",
"",
"",
"2018068650",
"2221298318",
"2950966695"
],
"abstract": [
"We address the problem of action detection in videos. Driven by the latest progress in object detection from 2D images, we build action models using rich feature hierarchies derived from shape and kinematic cues. We incorporate appearance and motion in two ways. First, starting from image region proposals we select those that are motion salient and thus are more likely to contain the action. This leads to a significant reduction in the number of regions being processed and allows for faster computations. Second, we extract spatio-temporal feature representations to build strong classifiers using Convolutional Neural Networks. We link our predictions to produce detections consistent in time, which we call action tubes. We show that our approach outperforms other techniques in the task of action detection.",
"This paper contributes to automatic classification and localization of human actions in video. Whereas motion is the key ingredient in modern approaches, we assess the benefits of having objects in the video representation. Rather than considering a handful of carefully selected and localized objects, we conduct an empirical study on the benefit of encoding 15,000 object categories for action using 6 datasets totaling more than 200 hours of video and covering 180 action classes. Our key contributions are i) the first in-depth study of encoding objects for actions, ii) we show that objects matter for actions, and are often semantically relevant as well. iii) We establish that actions have object preferences. Rather than using all objects, selection is advantageous for action recognition. iv)We reveal that object-action relations are generic, which allows to transferring these relationships from the one domain to the other. And, v) objects, when combined with motion, improve the state-of-the-art for both action classification and localization.",
"This paper describes a discriminatively trained, multiscale, deformable part model for object detection. Our system achieves a two-fold improvement in average precision over the best performance in the 2006 PASCAL person detection challenge. It also outperforms the best results in the 2007 challenge in ten out of twenty categories. The system relies heavily on deformable parts. While deformable part models have become quite popular, their value had not been demonstrated on difficult benchmarks such as the PASCAL challenge. Our system also relies heavily on new methods for discriminative training. We combine a margin-sensitive approach for data mining hard negative examples with a formalism we call latent SVM. A latent SVM, like a hidden CRF, leads to a non-convex training problem. However, a latent SVM is semi-convex and the training problem becomes convex once latent information is specified for the positive examples. We believe that our training methods will eventually make possible the effective use of more latent information such as hierarchical (grammar) models and models involving latent three dimensional pose.",
"Deformable part models have achieved impressive performance for object detection, even on difficult image datasets. This paper explores the generalization of deformable part models from 2D images to 3D spatiotemporal volumes to better study their effectiveness for action detection in video. Actions are treated as spatiotemporal patterns and a deformable part model is generated for each action from a collection of examples. For each action model, the most discriminative 3D sub volumes are automatically selected as parts and the spatiotemporal relations between their locations are learned. By focusing on the most distinctive parts of each action, our models adapt to intra-class variation and show robustness to clutter. Extensive experiments on several video datasets demonstrate the strength of spatiotemporal DPMs for classifying and localizing actions.",
"Real-world actions occur often in crowded, dynamic environments. This poses a difficult challenge for current approaches to video event detection because it is difficult to segment the actor from the background due to distracting motion from other objects in the scene. We propose a technique for event recognition in crowded videos that reliably identifies actions in the presence of partial occlusion and background clutter. Our approach is based on three key ideas: (1) we efficiently match the volumetric representation of an event against oversegmented spatio-temporal video volumes; (2) we augment our shape-based features using flow; (3) rather than treating an event template as an atomic entity, we separately match by parts (both in space and time), enabling robustness against occlusions and actor variability. Our experiments on human actions, such as picking up a dropped object or waving in a crowd show reliable detection with few false positives.",
"",
"",
"This paper considers the problem of action localization, where the objective is to determine when and where certain actions appear. We introduce a sampling strategy to produce 2D+t sequences of bounding boxes, called tubelets. Compared to state-of-the-art alternatives, this drastically reduces the number of hypotheses that are likely to include the action of interest. Our method is inspired by a recent technique introduced in the context of image localization. Beyond considering this technique for the first time for videos, we revisit this strategy for 2D+t sequences obtained from super-voxels. Our sampling strategy advantageously exploits a criterion that reflects how action related motion deviates from background motion. We demonstrate the interest of our approach by extensive experiments on two public datasets: UCF Sports and MSR-II. Our approach significantly outperforms the state-of-the-art on both datasets, while restricting the search of actions to a fraction of possible bounding box sequences.",
"This paper presents an efficient approach for localizing actions by learning contextual relations, in the form of relative locations between different video regions. We begin by over-segmenting the videos into supervoxels, which have the ability to preserve action boundaries and also reduce the complexity of the problem. Context relations are learned during training which capture displacements from all the supervoxels in a video to those belonging to foreground actions. Then, given a testing video, we select a supervoxel randomly and use the context information acquired during training to estimate the probability of each supervoxel belonging to the foreground action. The walk proceeds to a new supervoxel and the process is repeated for a few steps. This \"context walk\" generates a conditional distribution of an action over all the supervoxels. A Conditional Random Field is then used to find action proposals in the video, whose confidences are obtained using SVMs. We validated the proposed approach on several datasets and show that context in the form of relative displacements between supervoxels can be extremely useful for action localization. This also results in significantly fewer evaluations of the classifier, in sharp contrast to the alternate sliding window approaches.",
"We propose an effective approach for spatio-temporal action localization in realistic videos. The approach first detects proposals at the frame-level and scores them with a combination of static and motion CNN features. It then tracks high-scoring proposals throughout the video using a tracking-by-detection approach. Our tracker relies simultaneously on instance-level and class-level detectors. The tracks are scored using a spatio-temporal motion histogram, a descriptor at the track level, in combination with the CNN features. Finally, we perform temporal localization of the action using a sliding-window approach at the track level. We present experimental results for spatio-temporal localization on the UCF-Sports, J-HMDB and UCF-101 action localization datasets, where our approach outperforms the state of the art with a margin of 15 , 7 and 12 respectively in mAP."
]
} |
1703.10664 | 2604856029 | Deep learning has been demonstrated to achieve excellent results for image classification and object detection. However, the impact of deep learning on video analysis (e.g. action detection and recognition) has been limited due to complexity of video data and lack of annotations. Previous convolutional neural networks (CNN) based video action detection approaches usually consist of two major steps: frame-level action proposal detection and association of proposals across frames. Also, these methods employ two-stream CNN framework to handle spatial and temporal feature separately. In this paper, we propose an end-to-end deep network called Tube Convolutional Neural Network (T-CNN) for action detection in videos. The proposed architecture is a unified network that is able to recognize and localize action based on 3D convolution features. A video is first divided into equal length clips and for each clip a set of tube proposals are generated next based on 3D Convolutional Network (ConvNet) features. Finally, the tube proposals of different clips are linked together employing network flow and spatio-temporal action detection is performed using these linked video proposals. Extensive experiments on several video datasets demonstrate the superior performance of T-CNN for classifying and localizing actions in both trimmed and untrimmed videos compared to state-of-the-arts. | For object detection in images, Girshick al propose Region CNN (R-CNN) @cite_15 . In their approach region proposals are extracted using selective search. Then the candidate regions are warped to a fixed size and fed into ConvNet to extract CNN features. Finally, SVM model is trained for object classification. A fast version of R-CNN, Fast R-CNN, is presented in @cite_23 . Compared to the multi-stage pipeline of R-CNN, fast R-CNN incorporates object classifier in the network and trains object classifier and bounding box regressor simultaneously. Region of interest (RoI) pooling layer is introduced to extract fixed-length feature vectors for bounding boxes with different sizes. Recently, faster R-CNN is proposed in @cite_16 . It introduces a RPN (Region Proposal Network) to replace selective search for proposal generation. RPN shares full image convolutional features with the detection network, thus the proposal generation is almost cost-free. Faster R-CNN achieves state-of-the-art object detection performance while being efficient during testing. Motivated by its high performance, in this paper we explore generalizing faster R-CNN from 2D image regions to 3D video volumes for action detection. | {
"cite_N": [
"@cite_15",
"@cite_16",
"@cite_23"
],
"mid": [
"2102605133",
"2953106684",
""
],
"abstract": [
"Object detection performance, as measured on the canonical PASCAL VOC dataset, has plateaued in the last few years. The best-performing methods are complex ensemble systems that typically combine multiple low-level image features with high-level context. In this paper, we propose a simple and scalable detection algorithm that improves mean average precision (mAP) by more than 30 relative to the previous best result on VOC 2012 -- achieving a mAP of 53.3 . Our approach combines two key insights: (1) one can apply high-capacity convolutional neural networks (CNNs) to bottom-up region proposals in order to localize and segment objects and (2) when labeled training data is scarce, supervised pre-training for an auxiliary task, followed by domain-specific fine-tuning, yields a significant performance boost. Since we combine region proposals with CNNs, we call our method R-CNN: Regions with CNN features. We also present experiments that provide insight into what the network learns, revealing a rich hierarchy of image features. Source code for the complete system is available at http: www.cs.berkeley.edu rbg rcnn.",
"State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN 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. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features---using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model, 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, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks. Code has been made publicly available.",
""
]
} |
1703.10667 | 2605373614 | Recent two-stream deep Convolutional Neural Networks (ConvNets) have made significant progress in recognizing human actions in videos. Despite their success, methods extending the basic two-stream ConvNet have not systematically explored possible network architectures to further exploit spatiotemporal dynamics within video sequences. Further, such networks often use different baseline two-stream networks. Therefore, the differences and the distinguishing factors between various methods using Recurrent Neural Networks (RNN) or convolutional networks on temporally-constructed feature vectors (Temporal-ConvNet) are unclear. In this work, we first demonstrate a strong baseline two-stream ConvNet using ResNet-101. We use this baseline to thoroughly examine the use of both RNNs and Temporal-ConvNets for extracting spatiotemporal information. Building upon our experimental results, we then propose and investigate two different networks to further integrate spatiotemporal information: 1) temporal segment RNN and 2) Inception-style Temporal-ConvNet. We demonstrate that using both RNNs (using LSTMs) and Temporal-ConvNets on spatiotemporal feature matrices are able to exploit spatiotemporal dynamics to improve the overall performance. However, each of these methods require proper care to achieve state-of-the-art performance; for example, LSTMs require pre-segmented data or else they cannot fully exploit temporal information. Our analysis identifies specific limitations for each method that could form the basis of future work. Our experimental results on UCF101 and HMDB51 datasets achieve state-of-the-art performances, 94.1 and 69.0 , respectively, without requiring extensive temporal augmentation. | Instead of integrating temporal information via 3D convolutional kernels, @cite_15 fed spatial features extracted from each time step to a recurrent network with LSTM cells. In contrast to the traditional models which can only take a fixed number of temporal inputs and have limited spatiotemporal receptive fields, the proposed Long-term Recurrent Convolutional Networks (LRCN) can directly take variable length inputs and learn long-term dependencies. | {
"cite_N": [
"@cite_15"
],
"mid": [
"2951183276"
],
"abstract": [
"Models based on deep convolutional networks have dominated recent image interpretation tasks; we investigate whether models which are also recurrent, or \"temporally deep\", are effective for tasks involving sequences, visual and otherwise. We develop a novel recurrent convolutional architecture suitable for large-scale visual learning which is end-to-end trainable, and demonstrate the value of these models on benchmark video recognition tasks, image description and retrieval problems, and video narration challenges. In contrast to current models which assume a fixed spatio-temporal receptive field or simple temporal averaging for sequential processing, recurrent convolutional models are \"doubly deep\"' in that they can be compositional in spatial and temporal \"layers\". Such models may have advantages when target concepts are complex and or training data are limited. Learning long-term dependencies is possible when nonlinearities are incorporated into the network state updates. Long-term RNN models are appealing in that they directly can map variable-length inputs (e.g., video frames) to variable length outputs (e.g., natural language text) and can model complex temporal dynamics; yet they can be optimized with backpropagation. Our recurrent long-term models are directly connected to modern visual convnet models and can be jointly trained to simultaneously learn temporal dynamics and convolutional perceptual representations. Our results show such models have distinct advantages over state-of-the-art models for recognition or generation which are separately defined and or optimized."
]
} |
1703.10519 | 2604641614 | An energy harvesting (EH) transmitter communicating over a time-correlated wireless channel is considered. The transmitter is capable of sensing the current channel state, albeit at the cost of both energy and transmission time. The EH transmitter aims to maximize its long-term throughput by choosing one of the following actions: 1) defer its transmission to save energy for future use; 2) transmit reliably at a low rate; 3) transmit at a high rate; and 4) sense the channel to reveal the channel state at a cost of energy and transmission time, and then decide to defer or to transmit. The problem is formulated as a partially observable Markov decision process with a belief on the channel state. The optimal policy is shown to exhibit a threshold behavior on the belief state, with battery-dependent threshold values. The optimal threshold values and performance are characterized numerically via the value iteration algorithm as well as a policy search algorithm that exploits the threshold structure of the optimal policy. Our results demonstrate that, despite the associated time and energy cost, sensing the channel intelligently to track the channel state improves the achievable long-term throughput significantly as compared to the performance of those protocols lacking this ability as well as the one that always senses the channel. | Gilbert-Elliott channel model has been previously investigated in the context of scheduling an EH transmitter in @cite_25 , where the transmitter always has perfect CSI, obtained by sensing at every time slot. The transmitter makes a decision to defer or to transmit based on the current CSI and the battery state. Similarly, without considering the channel sensing capability, @cite_11 addresses the problem of optimal power management for an EH sensor over a multi-state wireless channel with memory. Unlike previous work, we take into account the energy cost of channel sensing which can be significant for a low-power EH transmitter. Therefore, in order to minimize the energy consumed for channel sensing, an EH transmitter does not necessarily sense the channel at every time slot, and instead, it keeps an updated belief of the channel state according to its past observations, and only occasionally senses the current channel state. | {
"cite_N": [
"@cite_25",
"@cite_11"
],
"mid": [
"2094169594",
"2005435577"
],
"abstract": [
"In this paper, we consider a source node that operates over a time varying channel with energy harvesting capability. The goal of the source is to maximize the average number of successfully delivered packets per time slot. The source is able to choose whether to transmit a packet or defer the transmission in each time slot. The decision which is chosen by the source depends on the channel information available and the length of the energy queue. We formulate the problem of finding the optimal policy as a Markovian decision problem. We show some properties of the value function that represents the discounted number of successfully delivered packets per time slot. We prove that the optimal policy is a threshold type policy depending on the state of the channel and the length of the energy queue. We also derive an upper bound for the average number of packets per time slots successfully received by the destination. We show using numerical results that this bound is a tight bound on the performance of the optimal policy. And we consider the case of time varying channel but without channel state information (CSI). Then, we study the impact of channel time varying nature and the availability of CSI. In this case, we show that the optimal policy is a greedy policy. The performance of this greedy policy is also calculated.",
"This paper addresses the problem of finding outage- optimal power control policies for wireless energy harvesting sensor (EHS) nodes with automatic repeat request (ARQ)-based packet transmissions. The power control policy of the EHS specifies the transmission power for each packet transmission attempt, based on all the information available at the EHS. In particular, the acknowledgement (ACK) or negative acknowledgement (NACK) messages received provide the EHS with partial information about the channel state. We solve the problem of finding an optimal power control policy by casting it as a partially observable Markov decision process (POMDP). We study the structure of the optimal power policy in two ways. First, for the special case of binary power levels at the EHS, we show that the optimal policy for the underlying Markov decision process (MDP) when the channel state is observable is a threshold policy in the battery state. Second, we benchmark the performance of the EHS by rigorously analyzing the outage probability of a general fixed-power transmission scheme, where the EHS uses a predetermined power level at each slot within the frame. Monte Carlo simulation results illustrate the performance of the POMDP approach and verify the accuracy of the analysis. They also show that the POMDP solutions can significantly outperform conventional ad hoc approaches."
]
} |
1703.10519 | 2604641614 | An energy harvesting (EH) transmitter communicating over a time-correlated wireless channel is considered. The transmitter is capable of sensing the current channel state, albeit at the cost of both energy and transmission time. The EH transmitter aims to maximize its long-term throughput by choosing one of the following actions: 1) defer its transmission to save energy for future use; 2) transmit reliably at a low rate; 3) transmit at a high rate; and 4) sense the channel to reveal the channel state at a cost of energy and transmission time, and then decide to defer or to transmit. The problem is formulated as a partially observable Markov decision process with a belief on the channel state. The optimal policy is shown to exhibit a threshold behavior on the belief state, with battery-dependent threshold values. The optimal threshold values and performance are characterized numerically via the value iteration algorithm as well as a policy search algorithm that exploits the threshold structure of the optimal policy. Our results demonstrate that, despite the associated time and energy cost, sensing the channel intelligently to track the channel state improves the achievable long-term throughput significantly as compared to the performance of those protocols lacking this ability as well as the one that always senses the channel. | Channel sensing is an essential part of opportunistic and cognitive spectrum access. In @cite_0 , the authors investigate the problem of optimal access to a Gilbert-Elliot channel, wherein an energy-unlimited transmitter senses the channel at every time slot. In @cite_17 channel sensing is done only occasionally. The transmitter can decide to transmit at a high or a low rate without sensing the channel, or it can first sense the channel and transmit at a reduced rate due to the time spent for sensing. However, the energy cost of sensing is ignored in @cite_17 . Energy cost of channel sensing has been previously studied in @cite_20 for a multiple-input single-output fading channel without memory when both the transmitter and the receiver harvest energy. | {
"cite_N": [
"@cite_0",
"@cite_20",
"@cite_17"
],
"mid": [
"2142819538",
"1992919284",
"2152337784"
],
"abstract": [
"We propose decentralized cognitive MAC protocols that allow secondary users to independently search for spectrum opportunities without a central coordinator or a dedicated communication channel. Recognizing hardware and energy constraints, we assume that a secondary user may not be able to perform full-spectrum sensing or may not be willing to monitor the spectrum when it has no data to transmit. We develop an analytical framework for opportunistic spectrum access based on the theory of partially observable Markov decision process (POMDP). This decision-theoretic approach integrates the design of spectrum access protocols at the MAC layer with spectrum sensing at the physical layer and traffic statistics determined by the application layer of the primary network. It also allows easy incorporation of spectrum sensing error and constraint on the probability of colliding with the primary users. Under this POMDP framework, we propose cognitive MAC protocols that optimize the performance of secondary users while limiting the interference perceived by primary users. A suboptimal strategy with reduced complexity yet comparable performance is developed. Without additional control message exchange between the secondary transmitter and receiver, the proposed decentralized protocols ensure synchronous hopping in the spectrum between the transmitter and the receiver in the presence of collisions and spectrum sensing errors",
"Optimization of a point-to-point (p2p) multiple-input single-output (MISO) communication system is considered when both the transmitter (TX) and the receiver (RX) have energy harvesting (EH) capabilities. The RX is interested in feeding back the channel state information (CSI) to the TX to help improve the transmission rate. The objective is to maximize the throughput by a deadline, subject to the EH constraints at the TX and the RX. The throughput metric considered is an upper bound on the ergodic rate of the MISO channel with beamforming and limited feedback. Feedback bit allocation and transmission policies that maximize the upper bound on the ergodic rate are obtained. Tools from majorization theory are used to simplify the formulated optimization problems. Optimal policies obtained for the modified problem outperform the naive scheme in which no intelligent management of energy is performed.",
"In this paper a communication system operating over a Gilbert-Elliot channel is studied. The goal of the transmitter is to maximize the number of successfully transmitted bits. This is achieved by choosing among three possible actions: (i) betting aggressively by using a weak code that allows transmission with a high data rate but provides no protection against a bad channel, ii) betting conservatively by using a strong code that perfectly protects the communication against a bad channel but does not allow a high data rate, iii) betting opportunistically by sensing the channel for a fixed duration and then deciding which code to use. The problem is formulated and solved using the theory of Markov decision processes (MDPs). It is shown that the optimal strategy has a simple threshold structure. Closed form expressions and simplified procedures for the computation of the threshold policies in terms of the system parameters are provided."
]
} |
1703.10155 | 2604433135 | We present variational generative adversarial networks, a general learning framework that combines a variational auto-encoder with a generative adversarial network, for synthesizing images in fine-grained categories, such as faces of a specific person or objects in a category. Our approach models an image as a composition of label and latent attributes in a probabilistic model. By varying the fine-grained category label fed into the resulting generative model, we can generate images in a specific category with randomly drawn values on a latent attribute vector. Our approach has two novel aspects. First, we adopt a cross entropy loss for the discriminative and classifier network, but a mean discrepancy objective for the generative network. This kind of asymmetric loss function makes the GAN training more stable. Second, we adopt an encoder network to learn the relationship between the latent space and the real image space, and use pairwise feature matching to keep the structure of generated images. We experiment with natural images of faces, flowers, and birds, and demonstrate that the proposed models are capable of generating realistic and diverse samples with fine-grained category labels. We further show that our models can be applied to other tasks, such as image inpainting, super-resolution, and data augmentation for training better face recognition models. | Conventional wisdom and early research of generative models, including Principle Component Analysis (PCA) @cite_22 , Independent Component Analysis (ICA) @cite_44 ,and the Gaussian Mixture Model (GMM) @cite_25 @cite_26 @cite_27 , all assume a simple formation of data. They have difficulty modeling complex patterns of irregular distributions. Later works, such as the Hidden Markov Model (HMM) @cite_15 , Markov Random Field (MRF) @cite_7 , and restricted Boltzmann machines (RBMs) @cite_29 @cite_8 , discriminatively train generative models @cite_47 , limiting their results on texture patches, digital numbers or well aligned faces, due to a lack of effective feature representations. | {
"cite_N": [
"@cite_47",
"@cite_26",
"@cite_22",
"@cite_7",
"@cite_8",
"@cite_29",
"@cite_44",
"@cite_27",
"@cite_15",
"@cite_25"
],
"mid": [
"2163176424",
"2113447708",
"2098693229",
"2121331909",
"189596042",
"2100495367",
"",
"2079256176",
"2157548127",
"2134199473"
],
"abstract": [
"Generative model learning is one of the key problems in machine learning and computer vision. Currently the use of generative models is limited due to the difficulty in effectively learning them. A new learning framework is proposed in this paper which progressively learns a target generative distribution through discriminative approaches. This framework provides many interesting aspects to the literature. From the generative model side: (1) A reference distribution is used to assist the learning process, which removes the need for a sampling processes in the early stages. (2) The classification power of discriminative approaches, e.g. boosting, is directly utilized. (3) The ability to select explore features from a large candidate pool allows us to make nearly no assumptions about the training data. From the discriminative model side: (1) This framework improves the modeling capability of discriminative models. (2) It can start with source training data only and gradually \"invent\" negative samples. (3) We show how sampling schemes can be introduced to discriminative models. (4) The learning procedure helps to tighten the decision boundaries for classification, and therefore, improves robustness. In this paper, we show a variety of applications including texture modeling and classification, non-photorealistic rendering, learning image statistics denoising, and face modeling. The framework handles both homogeneous patterns, e.g. textures, and inhomogeneous patterns, e.g. faces, with nearly an identical parameter setting for all the tasks in the learning stage.",
"We introduce Gaussian mixture models of 'structure' and colour features in order to classify coloured textures in images, with a view to the retrieval of textured colour images from databases. Classifications are performed separately using structure and colour and then combined using a confidence criterion. We apply the models to the VisTex database and to the classification of man-made and natural areas in aerial images. We compare these models with others in the literature, and show an overall improvement in performance.",
"An approach to the detection and identification of human faces is presented, and a working, near-real-time face recognition system which tracks a subject's head and then recognizes the person by comparing characteristics of the face to those of known individuals is described. This approach treats face recognition as a two-dimensional recognition problem, taking advantage of the fact that faces are normally upright and thus may be described by a small set of 2-D characteristic views. Face images are projected onto a feature space ('face space') that best encodes the variation among known face images. The face space is defined by the 'eigenfaces', which are the eigenvectors of the set of faces; they do not necessarily correspond to isolated features such as eyes, ears, and noses. The framework provides the ability to learn to recognize new faces in an unsupervised manner. >",
"Probabilistic models of natural images are usually evaluated by measuring performance on rather indirect tasks, such as denoising and inpainting. A more direct way to evaluate a generative model is to draw samples from it and to check whether statistical properties of the samples match the statistics of natural images. This method is seldom used with high-resolution images, because current models produce samples that are very different from natural images, as assessed by even simple visual inspection. We investigate the reasons for this failure and we show that by augmenting existing models so that there are two sets of latent variables, one set modelling pixel intensities and the other set modelling image-specific pixel covariances, we are able to generate high-resolution images that look much more realistic than before. The overall model can be interpreted as a gated MRF where both pair-wise dependencies and mean intensities of pixels are modulated by the states of latent variables. Finally, we confirm that if we disallow weight-sharing between receptive fields that overlap each other, the gated MRF learns more efficient internal representations, as demonstrated in several recognition tasks.",
"We present a new learning algorithm for Boltzmann machines that contain many layers of hidden variables. Data-dependent expectations are estimated using a variational approximation that tends to focus on a single mode, and dataindependent expectations are approximated using persistent Markov chains. The use of two quite different techniques for estimating the two types of expectation that enter into the gradient of the log-likelihood makes it practical to learn Boltzmann machines with multiple hidden layers and millions of parameters. The learning can be made more efficient by using a layer-by-layer “pre-training” phase that allows variational inference to be initialized with a single bottomup pass. We present results on the MNIST and NORB datasets showing that deep Boltzmann machines learn good generative models and perform well on handwritten digit and visual object recognition tasks.",
"High-dimensional data can be converted to low-dimensional codes by training a multilayer neural network with a small central layer to reconstruct high-dimensional input vectors. Gradient descent can be used for fine-tuning the weights in such “autoencoder” networks, but this works well only if the initial weights are close to a good solution. We describe an effective way of initializing the weights that allows deep autoencoder networks to learn low-dimensional codes that work much better than principal components analysis as a tool to reduce the dimensionality of data.",
"",
"We present a probabilistic model for natural images that is based on mixtures of Gaussian scale mixtures and a simple multiscale representation. We show that it is able to generate images with interesting higher-order correlations when trained on natural images or samples from an occlusion-based model. More importantly, our multiscale model allows for a principled evaluation. While it is easy to generate visually appealing images, we demonstrate that our model also yields the best performance reported to date when evaluated with respect to the cross-entropy rate, a measure tightly linked to the average log-likelihood. The ability to quantitatively evaluate our model differentiates it from other multiscale models, for which evaluation of these kinds of measures is usually intractable.",
"Hidden Markov models (HMMs) have been used prominently and successfully in speech recognition and, more recently, in handwriting recognition. Consequently, they seem ideal for visual recognition of complex, structured hand gestures such as are found in sign language. We describe a real-time HMM-based system for recognizing sentence level American Sign Language (ASL) which attains a word accuracy of 99.2 without explicitly modeling the fingers.",
"We build up the mathematical connection between the “Expectation-Maximization” (EM) algorithm and gradient-based approaches for maximum likelihood learning of finite gaussian mixtures. We show that the EM step in parameter space is obtained from the gradient via a projection matrix P, and we provide an explicit expression for the matrix. We then analyze the convergence of EM in terms of special properties of P and provide new results analyzing the effect that P has on the likelihood surface. Based on these mathematical results, we present a comparative discussion of the advantages and disadvantages of EM and other algorithms for the learning of gaussian mixture models."
]
} |
1703.10155 | 2604433135 | We present variational generative adversarial networks, a general learning framework that combines a variational auto-encoder with a generative adversarial network, for synthesizing images in fine-grained categories, such as faces of a specific person or objects in a category. Our approach models an image as a composition of label and latent attributes in a probabilistic model. By varying the fine-grained category label fed into the resulting generative model, we can generate images in a specific category with randomly drawn values on a latent attribute vector. Our approach has two novel aspects. First, we adopt a cross entropy loss for the discriminative and classifier network, but a mean discrepancy objective for the generative network. This kind of asymmetric loss function makes the GAN training more stable. Second, we adopt an encoder network to learn the relationship between the latent space and the real image space, and use pairwise feature matching to keep the structure of generated images. We experiment with natural images of faces, flowers, and birds, and demonstrate that the proposed models are capable of generating realistic and diverse samples with fine-grained category labels. We further show that our models can be applied to other tasks, such as image inpainting, super-resolution, and data augmentation for training better face recognition models. | There have been many recent developments of deep generative models @cite_23 @cite_36 @cite_13 @cite_21 @cite_34 @cite_24 @cite_3 @cite_43 @cite_12 . Since deep hierarchical architectures allow them to capture complex structures in the data, all these methods show promising results in generating natural images that are far more realistic than conventional generative models. Among them are three main themes: Variational Auto-encoder (VAE) @cite_36 @cite_21 , Generative Adversarial Network (GAN) @cite_13 @cite_34 @cite_43 , and Autoregression @cite_23 . | {
"cite_N": [
"@cite_36",
"@cite_21",
"@cite_3",
"@cite_24",
"@cite_43",
"@cite_23",
"@cite_34",
"@cite_13",
"@cite_12"
],
"mid": [
"",
"1909320841",
"2951523806",
"2202109488",
"2432004435",
"2135181320",
"2173520492",
"",
"2950629931"
],
"abstract": [
"",
"We marry ideas from deep neural networks and approximate Bayesian inference to derive a generalised class of deep, directed generative models, endowed with a new algorithm for scalable inference and learning. Our algorithm introduces a recognition model to represent approximate posterior distributions, and that acts as a stochastic encoder of the data. We develop stochastic back-propagation -- rules for back-propagation through stochastic variables -- and use this to develop an algorithm that allows for joint optimisation of the parameters of both the generative and recognition model. We demonstrate on several real-world data sets that the model generates realistic samples, provides accurate imputations of missing data and is a useful tool for high-dimensional data visualisation.",
"In this paper we introduce a generative parametric model capable of producing high quality samples of natural images. Our approach uses a cascade of convolutional networks within a Laplacian pyramid framework to generate images in a coarse-to-fine fashion. At each level of the pyramid, a separate generative convnet model is trained using the Generative Adversarial Nets (GAN) approach (). Samples drawn from our model are of significantly higher quality than alternate approaches. In a quantitative assessment by human evaluators, our CIFAR10 samples were mistaken for real images around 40 of the time, compared to 10 for samples drawn from a GAN baseline model. We also show samples from models trained on the higher resolution images of the LSUN scene dataset.",
"We present an autoencoder that leverages learned representations to better measure similarities in data space. By combining a variational autoencoder with a generative adversarial network we can use learned feature representations in the GAN discriminator as basis for the VAE reconstruction objective. Thereby, we replace element-wise errors with feature-wise errors to better capture the data distribution while offering invariance towards e.g. translation. We apply our method to images of faces and show that it outperforms VAEs with element-wise similarity measures in terms of visual fidelity. Moreover, we show that the method learns an embedding in which high-level abstract visual features (e.g. wearing glasses) can be modified using simple arithmetic.",
"We present a variety of new architectural features and training procedures that we apply to the generative adversarial networks (GANs) framework. We focus on two applications of GANs: semi-supervised learning, and the generation of images that humans find visually realistic. Unlike most work on generative models, our primary goal is not to train a model that assigns high likelihood to test data, nor do we require the model to be able to learn well without using any labels. Using our new techniques, we achieve state-of-the-art results in semi-supervised classification on MNIST, CIFAR-10 and SVHN. The generated images are of high quality as confirmed by a visual Turing test: our model generates MNIST samples that humans cannot distinguish from real data, and CIFAR-10 samples that yield a human error rate of 21.3 . We also present ImageNet samples with unprecedented resolution and show that our methods enable the model to learn recognizable features of ImageNet classes.",
"We describe a new approach for modeling the distribution of high-dimensional vectors of discrete variables. This model is inspired by the restricted Boltzmann machine (RBM), which has been shown to be a powerful model of such distributions. However, an RBM typically does not provide a tractable distribution estimator, since evaluating the probability it assigns to some given observation requires the computation of the so-called partition function, which itself is intractable for RBMs of even moderate size. Our model circumvents this diculty by decomposing the joint distribution of observations into tractable conditional distributions and modeling each conditional using a non-linear function similar to a conditional of an RBM. Our model can also be interpreted as an autoencoder wired such that its output can be used to assign valid probabilities to observations. We show that this new model outperforms other multivariate binary distribution estimators on several datasets and performs similarly to a large (but intractable) RBM.",
"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 train generative 'up-convolutional' neural networks which are able to generate images of objects given object style, viewpoint, and color. We train the networks on rendered 3D models of chairs, tables, and cars. Our experiments show that the networks do not merely learn all images by heart, but rather find a meaningful representation of 3D models allowing them to assess the similarity of different models, interpolate between given views to generate the missing ones, extrapolate views, and invent new objects not present in the training set by recombining training instances, or even two different object classes. Moreover, we show that such generative networks can be used to find correspondences between different objects from the dataset, outperforming existing approaches on this task."
]
} |
1703.10155 | 2604433135 | We present variational generative adversarial networks, a general learning framework that combines a variational auto-encoder with a generative adversarial network, for synthesizing images in fine-grained categories, such as faces of a specific person or objects in a category. Our approach models an image as a composition of label and latent attributes in a probabilistic model. By varying the fine-grained category label fed into the resulting generative model, we can generate images in a specific category with randomly drawn values on a latent attribute vector. Our approach has two novel aspects. First, we adopt a cross entropy loss for the discriminative and classifier network, but a mean discrepancy objective for the generative network. This kind of asymmetric loss function makes the GAN training more stable. Second, we adopt an encoder network to learn the relationship between the latent space and the real image space, and use pairwise feature matching to keep the structure of generated images. We experiment with natural images of faces, flowers, and birds, and demonstrate that the proposed models are capable of generating realistic and diverse samples with fine-grained category labels. We further show that our models can be applied to other tasks, such as image inpainting, super-resolution, and data augmentation for training better face recognition models. | Generative Adversarial Network (GAN) @cite_13 @cite_34 @cite_43 is another popular generative model. It simultaneously trains two models: a generative model to synthesize samples, and a discriminative model to differentiate between natural and synthesized samples. However, the GAN model is hard to converge in the training stage and the samples generated from GAN are often far from natural. Recently, many works have tried to improve the quality of the generated samples. For example, the Wasserstein GAN (WGAN) @cite_35 uses Earth Mover Distance as an objective for training GANs, and McGAN @cite_38 uses mean and covariance feature matching. They need to limit the range of the parameters of the discriminator which will decrease discriminative power. Loss-Sensitive GAN @cite_18 learns a loss function which can quantify the quality of generated samples and uses this loss function to generate high-quality images. There are also methods which tried to combine GAN and VAE, e.g. , VAE GAN @cite_24 and adversarial autoencoders @cite_45 . They are closely related to and partly inspired our work. | {
"cite_N": [
"@cite_35",
"@cite_38",
"@cite_18",
"@cite_24",
"@cite_43",
"@cite_45",
"@cite_34",
"@cite_13"
],
"mid": [
"",
"2594154005",
"2580360036",
"2202109488",
"2432004435",
"",
"2173520492",
""
],
"abstract": [
"",
"We introduce new families of Integral Probability Metrics (IPM) for training Generative Adversarial Networks (GAN). Our IPMs are based on matching statistics of distributions embedded in a finite dimensional feature space. Mean and covariance feature matching IPMs allow for stable training of GANs, which we will call McGan. McGan minimizes a meaningful loss between distributions.",
"In this paper, we present the Lipschitz regularization theory and algorithms for a novel Loss-Sensitive Generative Adversarial Network (LS-GAN). Specifically, it trains a loss function to distinguish between real and fake samples by designated margins, while learning a generator alternately to produce realistic samples by minimizing their losses. The LS-GAN further regularizes its loss function with a Lipschitz regularity condition on the density of real data, yielding a regularized model that can better generalize to produce new data from a reasonable number of training examples than the classic GAN. We will further present a Generalized LS-GAN (GLS-GAN) and show it contains a large family of regularized GAN models, including both LS-GAN and Wasserstein GAN, as its special cases. Compared with the other GAN models, we will conduct experiments to show both LS-GAN and GLS-GAN exhibit competitive ability in generating new images in terms of the Minimum Reconstruction Error (MRE) assessed on a separate test set. We further extend the LS-GAN to a conditional form for supervised and semi-supervised learning problems, and demonstrate its outstanding performance on image classification tasks.",
"We present an autoencoder that leverages learned representations to better measure similarities in data space. By combining a variational autoencoder with a generative adversarial network we can use learned feature representations in the GAN discriminator as basis for the VAE reconstruction objective. Thereby, we replace element-wise errors with feature-wise errors to better capture the data distribution while offering invariance towards e.g. translation. We apply our method to images of faces and show that it outperforms VAEs with element-wise similarity measures in terms of visual fidelity. Moreover, we show that the method learns an embedding in which high-level abstract visual features (e.g. wearing glasses) can be modified using simple arithmetic.",
"We present a variety of new architectural features and training procedures that we apply to the generative adversarial networks (GANs) framework. We focus on two applications of GANs: semi-supervised learning, and the generation of images that humans find visually realistic. Unlike most work on generative models, our primary goal is not to train a model that assigns high likelihood to test data, nor do we require the model to be able to learn well without using any labels. Using our new techniques, we achieve state-of-the-art results in semi-supervised classification on MNIST, CIFAR-10 and SVHN. The generated images are of high quality as confirmed by a visual Turing test: our model generates MNIST samples that humans cannot distinguish from real data, and CIFAR-10 samples that yield a human error rate of 21.3 . We also present ImageNet samples with unprecedented resolution and show that our methods enable the model to learn recognizable features of ImageNet classes.",
"",
"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.",
""
]
} |
1703.10155 | 2604433135 | We present variational generative adversarial networks, a general learning framework that combines a variational auto-encoder with a generative adversarial network, for synthesizing images in fine-grained categories, such as faces of a specific person or objects in a category. Our approach models an image as a composition of label and latent attributes in a probabilistic model. By varying the fine-grained category label fed into the resulting generative model, we can generate images in a specific category with randomly drawn values on a latent attribute vector. Our approach has two novel aspects. First, we adopt a cross entropy loss for the discriminative and classifier network, but a mean discrepancy objective for the generative network. This kind of asymmetric loss function makes the GAN training more stable. Second, we adopt an encoder network to learn the relationship between the latent space and the real image space, and use pairwise feature matching to keep the structure of generated images. We experiment with natural images of faces, flowers, and birds, and demonstrate that the proposed models are capable of generating realistic and diverse samples with fine-grained category labels. We further show that our models can be applied to other tasks, such as image inpainting, super-resolution, and data augmentation for training better face recognition models. | VAEs and GANs can also be trained to conduct conditional generation, e.g. , CVAE @cite_14 and CGAN @cite_2 . By introducing additional conditionality, they can handle probabilistic one-to-many mapping problems. Recently there have been many interesting works based on CVAE and CGAN, including conditional face generation @cite_17 , Attribute2Image @cite_9 , text to image synthesis @cite_28 , forecasting from static images @cite_5 , and conditional image synthesis @cite_33 . All of them achieve impressive results. | {
"cite_N": [
"@cite_14",
"@cite_33",
"@cite_28",
"@cite_9",
"@cite_2",
"@cite_5",
"@cite_17"
],
"mid": [
"2188365844",
"2950776302",
"2949999304",
"2189246496",
"2125389028",
"2952390294",
""
],
"abstract": [
"Supervised deep learning has been successfully applied to many recognition problems. Although it can approximate a complex many-to-one function well when a large amount of training data is provided, it is still challenging to model complex structured output representations that effectively perform probabilistic inference and make diverse predictions. In this work, we develop a deep conditional generative model for structured output prediction using Gaussian latent variables. The model is trained efficiently in the framework of stochastic gradient variational Bayes, and allows for fast prediction using stochastic feed-forward inference. In addition, we provide novel strategies to build robust structured prediction algorithms, such as input noise-injection and multi-scale prediction objective at training. In experiments, we demonstrate the effectiveness of our proposed algorithm in comparison to the deterministic deep neural network counterparts in generating diverse but realistic structured output predictions using stochastic inference. Furthermore, the proposed training methods are complimentary, which leads to strong pixel-level object segmentation and semantic labeling performance on Caltech-UCSD Birds 200 and the subset of Labeled Faces in the Wild dataset.",
"Synthesizing high resolution photorealistic images has been a long-standing challenge in machine learning. In this paper we introduce new methods for the improved training of generative adversarial networks (GANs) for image synthesis. We construct a variant of GANs employing label conditioning that results in 128x128 resolution image samples exhibiting global coherence. We expand on previous work for image quality assessment to provide two new analyses for assessing the discriminability and diversity of samples from class-conditional image synthesis models. These analyses demonstrate that high resolution samples provide class information not present in low resolution samples. Across 1000 ImageNet classes, 128x128 samples are more than twice as discriminable as artificially resized 32x32 samples. In addition, 84.7 of the classes have samples exhibiting diversity comparable to real ImageNet data.",
"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 model- ing, 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.",
"This paper investigates a problem of generating images from visual attributes. Given the prevalent research for image recognition, the conditional image generation problem is relatively under-explored due to the challenges of learning a good generative model and handling rendering uncertainties in images. To address this, we propose a variety of attribute-conditioned deep variational auto-encoders that enjoy both effective representation learning and Bayesian modeling, from which images can be generated from specified attributes and sampled latent factors. We experiment with natural face images and demonstrate that the proposed models are capable of generating realistic faces with diverse appearance. We further evaluate the proposed models by performing attribute-conditioned image progression, transfer and retrieval. In particular, our generation method achieves superior performance in the retrieval experiment against traditional nearest-neighbor-based methods both qualitatively and quantitatively.",
"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.",
"In a given scene, humans can often easily predict a set of immediate future events that might happen. However, generalized pixel-level anticipation in computer vision systems is difficult because machine learning struggles with the ambiguity inherent in predicting the future. In this paper, we focus on predicting the dense trajectory of pixels in a scene, specifically what will move in the scene, where it will travel, and how it will deform over the course of one second. We propose a conditional variational autoencoder as a solution to this problem. In this framework, direct inference from the image shapes the distribution of possible trajectories, while latent variables encode any necessary information that is not available in the image. We show that our method is able to successfully predict events in a wide variety of scenes and can produce multiple different predictions when the future is ambiguous. Our algorithm is trained on thousands of diverse, realistic videos and requires absolutely no human labeling. In addition to non-semantic action prediction, we find that our method learns a representation that is applicable to semantic vision tasks.",
""
]
} |
1703.10155 | 2604433135 | We present variational generative adversarial networks, a general learning framework that combines a variational auto-encoder with a generative adversarial network, for synthesizing images in fine-grained categories, such as faces of a specific person or objects in a category. Our approach models an image as a composition of label and latent attributes in a probabilistic model. By varying the fine-grained category label fed into the resulting generative model, we can generate images in a specific category with randomly drawn values on a latent attribute vector. Our approach has two novel aspects. First, we adopt a cross entropy loss for the discriminative and classifier network, but a mean discrepancy objective for the generative network. This kind of asymmetric loss function makes the GAN training more stable. Second, we adopt an encoder network to learn the relationship between the latent space and the real image space, and use pairwise feature matching to keep the structure of generated images. We experiment with natural images of faces, flowers, and birds, and demonstrate that the proposed models are capable of generating realistic and diverse samples with fine-grained category labels. We further show that our models can be applied to other tasks, such as image inpainting, super-resolution, and data augmentation for training better face recognition models. | Generative ConvNet @cite_0 , demonstrates that a generative model can be derived from the commonly used discriminative ConvNet. @cite_39 and @cite_10 introduce a method that generates high quality images from features extracted from a trained classification model. PPGN @cite_46 performs exceptionally well in generating samples by using a gradient ascent and prior to the latent space of a generator. | {
"cite_N": [
"@cite_0",
"@cite_46",
"@cite_10",
"@cite_39"
],
"mid": [
"2949457404",
"2951140085",
"2963464195",
"2259643685"
],
"abstract": [
"We show that a generative random field model, which we call generative ConvNet, can be derived from the commonly used discriminative ConvNet, by assuming a ConvNet for multi-category classification and assuming one of the categories is a base category generated by a reference distribution. If we further assume that the non-linearity in the ConvNet is Rectified Linear Unit (ReLU) and the reference distribution is Gaussian white noise, then we obtain a generative ConvNet model that is unique among energy-based models: The model is piecewise Gaussian, and the means of the Gaussian pieces are defined by an auto-encoder, where the filters in the bottom-up encoding become the basis functions in the top-down decoding, and the binary activation variables detected by the filters in the bottom-up convolution process become the coefficients of the basis functions in the top-down deconvolution process. The Langevin dynamics for sampling the generative ConvNet is driven by the reconstruction error of this auto-encoder. The contrastive divergence learning of the generative ConvNet reconstructs the training images by the auto-encoder. The maximum likelihood learning algorithm can synthesize realistic natural image patterns.",
"Generating high-resolution, photo-realistic images has been a long-standing goal in machine learning. Recently, (2016) showed one interesting way to synthesize novel images by performing gradient ascent in the latent space of a generator network to maximize the activations of one or multiple neurons in a separate classifier network. In this paper we extend this method by introducing an additional prior on the latent code, improving both sample quality and sample diversity, leading to a state-of-the-art generative model that produces high quality images at higher resolutions (227x227) than previous generative models, and does so for all 1000 ImageNet categories. In addition, we provide a unified probabilistic interpretation of related activation maximization methods and call the general class of models \"Plug and Play Generative Networks\". PPGNs are composed of 1) a generator network G that is capable of drawing a wide range of image types and 2) a replaceable \"condition\" network C that tells the generator what to draw. We demonstrate the generation of images conditioned on a class (when C is an ImageNet or MIT Places classification network) and also conditioned on a caption (when C is an image captioning network). Our method also improves the state of the art of Multifaceted Feature Visualization, which generates the set of synthetic inputs that activate a neuron in order to better understand how deep neural networks operate. Finally, we show that our model performs reasonably well at the task of image inpainting. While image models are used in this paper, the approach is modality-agnostic and can be applied to many types of data.",
"Deep neural networks (DNNs) have demonstrated state-of-the-art results on many pattern recognition tasks, especially vision classification problems. Understanding the inner workings of such computational brains is both fascinating basic science that is interesting in its own right---similar to why we study the human brain---and will enable researchers to further improve DNNs. One path to understanding how a neural network functions internally is to study what each of its neurons has learned to detect. One such method is called activation maximization, which synthesizes an input (e.g. an image) that highly activates a neuron. Here we dramatically improve the qualitative state of the art of activation maximization by harnessing a powerful, learned prior: a deep generator network. The algorithm (1) generates qualitatively state-of-the-art synthetic images that look almost real, (2) reveals the features learned by each neuron in an interpretable way, (3) generalizes well to new datasets and somewhat well to different network architectures without requiring the prior to be relearned, and (4) can be considered as a high-quality generative method (in this case, by generating novel, creative, interesting, recognizable images).",
"Image-generating machine learning models are typically trained with loss functions based on distance in the image space. This often leads to over-smoothed results. We propose a class of loss functions, which we call deep perceptual similarity metrics (DeePSiM), that mitigate this problem. Instead of computing distances in the image space, we compute distances between image features extracted by deep neural networks. This metric better reflects perceptually similarity of images and thus leads to better results. We show three applications: autoencoder training, a modification of a variational autoencoder, and inversion of deep convolutional networks. In all cases, the generated images look sharp and resemble natural images."
]
} |
1703.10155 | 2604433135 | We present variational generative adversarial networks, a general learning framework that combines a variational auto-encoder with a generative adversarial network, for synthesizing images in fine-grained categories, such as faces of a specific person or objects in a category. Our approach models an image as a composition of label and latent attributes in a probabilistic model. By varying the fine-grained category label fed into the resulting generative model, we can generate images in a specific category with randomly drawn values on a latent attribute vector. Our approach has two novel aspects. First, we adopt a cross entropy loss for the discriminative and classifier network, but a mean discrepancy objective for the generative network. This kind of asymmetric loss function makes the GAN training more stable. Second, we adopt an encoder network to learn the relationship between the latent space and the real image space, and use pairwise feature matching to keep the structure of generated images. We experiment with natural images of faces, flowers, and birds, and demonstrate that the proposed models are capable of generating realistic and diverse samples with fine-grained category labels. We further show that our models can be applied to other tasks, such as image inpainting, super-resolution, and data augmentation for training better face recognition models. | Autoregression @cite_23 follows a different idea. It uses autoregressive connections to model images pixel by pixel. Its two variants, PixelRNN @cite_4 and PixelCNN @cite_31 , also produce excellent samples. | {
"cite_N": [
"@cite_31",
"@cite_4",
"@cite_23"
],
"mid": [
"2423557781",
"2953318193",
"2135181320"
],
"abstract": [
"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.",
"Modeling the distribution of natural images is a landmark problem in unsupervised learning. This task requires an image model that is at once expressive, tractable and scalable. We present a deep neural network that sequentially predicts the pixels in an image along the two spatial dimensions. Our method models the discrete probability of the raw pixel values and encodes the complete set of dependencies in the image. Architectural novelties include fast two-dimensional recurrent layers and an effective use of residual connections in deep recurrent networks. We achieve log-likelihood scores on natural images that are considerably better than the previous state of the art. Our main results also provide benchmarks on the diverse ImageNet dataset. Samples generated from the model appear crisp, varied and globally coherent.",
"We describe a new approach for modeling the distribution of high-dimensional vectors of discrete variables. This model is inspired by the restricted Boltzmann machine (RBM), which has been shown to be a powerful model of such distributions. However, an RBM typically does not provide a tractable distribution estimator, since evaluating the probability it assigns to some given observation requires the computation of the so-called partition function, which itself is intractable for RBMs of even moderate size. Our model circumvents this diculty by decomposing the joint distribution of observations into tractable conditional distributions and modeling each conditional using a non-linear function similar to a conditional of an RBM. Our model can also be interpreted as an autoencoder wired such that its output can be used to assign valid probabilities to observations. We show that this new model outperforms other multivariate binary distribution estimators on several datasets and performs similarly to a large (but intractable) RBM."
]
} |
1703.10106 | 2604906478 | We address human action recognition from multi-modal video data involving articulated pose and RGB frames and propose a two-stream approach. The pose stream is processed with a convolutional model taking as input a 3D tensor holding data from a sub-sequence. A specific joint ordering, which respects the topology of the human body, ensures that different convolutional layers correspond to meaningful levels of abstraction. The raw RGB stream is handled by a spatio-temporal soft-attention mechanism conditioned on features from the pose network. An LSTM network receives input from a set of image locations at each instant. A trainable glimpse sensor extracts features on a set of predefined locations specified by the pose stream, namely the 4 hands of the two people involved in the activity. Appearance features give important cues on hand motion and on objects held in each hand. We show that it is of high interest to shift the attention to different hands at different time steps depending on the activity itself. Finally a temporal attention mechanism learns how to fuse LSTM features over time. We evaluate the method on 3 datasets. State-of-the-art results are achieved on the largest dataset for human activity recognition, namely NTU-RGB+D, as well as on the SBU Kinect Interaction dataset. Performance close to state-of-the-art is achieved on the smaller MSR Daily Activity 3D dataset. | Activities, gestures and multimodal data Recent gesture action recognition methods dealing with several modalities typically process 2D+T RGB and or depth data as 3D. Sequences of frames are stacked into volumes and fed into convolutional layers at first stages @cite_47 @cite_7 @cite_17 @cite_9 @cite_15 . When additional pose data is available, the 3D joint positions are typically fed into a separate network. Preprocessing pose is reported to improve performance in some situations, e.g. augmenting coordinates with velocities and acceleration @cite_0 . Pose normalization (bone lengths and view point normalization) has been reported to help in certain situations @cite_40 . Fusing pose and raw video modalities is traditionally done as late fusion @cite_48 , or early through fusion layers @cite_23 . In @cite_44 , fusion strategies are learned together with model parameters with by stochastic regularization. | {
"cite_N": [
"@cite_7",
"@cite_15",
"@cite_48",
"@cite_9",
"@cite_0",
"@cite_44",
"@cite_40",
"@cite_23",
"@cite_47",
"@cite_17"
],
"mid": [
"1983364832",
"",
"",
"",
"1983592444",
"2580687518",
"2963192057",
"2295038166",
"28988658",
""
],
"abstract": [
"We consider the automated recognition of human actions in surveillance videos. Most current methods build classifiers based on complex handcrafted features computed from the raw inputs. Convolutional neural networks (CNNs) are a type of deep model that can act directly on the raw inputs. However, such models are currently limited to handling 2D inputs. In this paper, we develop a novel 3D CNN model for action recognition. This model extracts features from both the spatial and the temporal dimensions by performing 3D convolutions, thereby capturing the motion information encoded in multiple adjacent frames. The developed model generates multiple channels of information from the input frames, and the final feature representation combines information from all channels. To further boost the performance, we propose regularizing the outputs with high-level features and combining the predictions of a variety of different models. We apply the developed models to recognize human actions in the real-world environment of airport surveillance videos, and they achieve superior performance in comparison to baseline methods.",
"",
"",
"",
"Human action recognition under low observational latency is receiving a growing interest in computer vision due to rapidly developing technologies in human-robot interaction, computer gaming and surveillance. In this paper we propose a fast, simple, yet powerful non-parametric Moving Pose (MP) framework for low-latency human action and activity recognition. Central to our methodology is a moving pose descriptor that considers both pose information as well as differential quantities (speed and acceleration) of the human body joints within a short time window around the current frame. The proposed descriptor is used in conjunction with a modified kNN classifier that considers both the temporal location of a particular frame within the action sequence as well as the discrimination power of its moving pose descriptor compared to other frames in the training set. The resulting method is non-parametric and enables low-latency recognition, one-shot learning, and action detection in difficult unsegmented sequences. Moreover, the framework is real-time, scalable, and outperforms more sophisticated approaches on challenging benchmarks like MSR-Action3D or MSR-DailyActivities3D.",
"Model selection methods based on stochastic regularization such as Dropout have been widely used in deep learning due to their simplicity and effectiveness. The standard Dropout method treats all units, visible or hidden, in the same way, thus ignoring any information related to grouping or structure. Such structure is present in multi-modal learning applications such as affect analysis and gesture recognition, where subsets of units may correspond to individual modalities. In this paper we describe Modout, a model selection method based on stochastic regularization, which is particularly useful in the multi-modal setting. Different from previous methods, it is capable of learning whether or when to fuse two modalities in a layer, which is usually considered to be an architectural hyper-parameter by deep learning researchers and practitioners. Modout is evaluated on one synthetic and two real multi-modal datasets. The results indicate improved performance compared to other stochastic regularization methods. The result on the Montalbano dataset shows that learning a fusion structure by Modout is on par with a state-of-the-art carefully designed architecture.",
"We present a method for gesture detection and localisation based on multi-scale and multi-modal deep learning. Each visual modality captures spatial information at a particular spatial scale (such as motion of the upper body or a hand), and the whole system operates at three temporal scales. Key to our technique is a training strategy which exploits: i) careful initialization of individual modalities; and ii) gradual fusion involving random dropping of separate channels (dubbed ModDrop ) for learning cross-modality correlations while preserving uniqueness of each modality-specific representation. We present experiments on the ChaLearn 2014 Looking at People Challenge gesture recognition track, in which we placed first out of 17 teams. Fusing multiple modalities at several spatial and temporal scales leads to a significant increase in recognition rates, allowing the model to compensate for errors of the individual classifiers as well as noise in the separate channels. Futhermore, the proposed ModDrop training technique ensures robustness of the classifier to missing signals in one or several channels to produce meaningful predictions from any number of available modalities. In addition, we demonstrate the applicability of the proposed fusion scheme to modalities of arbitrary nature by experiments on the same dataset augmented with audio.",
"This paper describes a novel method called Deep Dynamic Neural Networks (DDNN) for multimodal gesture recognition. A semi-supervised hierarchical dynamic framework based on a Hidden Markov Model (HMM) is proposed for simultaneous gesture segmentation and recognition where skeleton joint information, depth and RGB images, are the multimodal input observations. Unlike most traditional approaches that rely on the construction of complex handcrafted features, our approach learns high-level spatio-temporal representations using deep neural networks suited to the input modality: a Gaussian-Bernouilli Deep Belief Network ( DBN ) to handle skelet al dynamics, and a 3D Convolutional Neural Network ( 3DCNN ) to manage and fuse batches of depth and RGB images. This is achieved through the modeling and learning of the emission probabilities of the HMM required to infer the gesture sequence. This purely data driven approach achieves a Jaccard index score of 0.81 in the ChaLearn LAP gesture spotting challenge. The performance is on par with a variety of state-of-the-art hand-tuned feature-based approaches and other learning-based methods, therefore opening the door to the use of deep learning techniques in order to further explore multimodal time series data.",
"We propose in this paper a fully automated deep model, which learns to classify human actions without using any prior knowledge. The first step of our scheme, based on the extension of Convolutional Neural Networks to 3D, automatically learns spatio-temporal features. A Recurrent Neural Network is then trained to classify each sequence considering the temporal evolution of the learned features for each timestep. Experimental results on the KTH dataset show that the proposed approach outperforms existing deep models, and gives comparable results with the best related works.",
""
]
} |
1703.10106 | 2604906478 | We address human action recognition from multi-modal video data involving articulated pose and RGB frames and propose a two-stream approach. The pose stream is processed with a convolutional model taking as input a 3D tensor holding data from a sub-sequence. A specific joint ordering, which respects the topology of the human body, ensures that different convolutional layers correspond to meaningful levels of abstraction. The raw RGB stream is handled by a spatio-temporal soft-attention mechanism conditioned on features from the pose network. An LSTM network receives input from a set of image locations at each instant. A trainable glimpse sensor extracts features on a set of predefined locations specified by the pose stream, namely the 4 hands of the two people involved in the activity. Appearance features give important cues on hand motion and on objects held in each hand. We show that it is of high interest to shift the attention to different hands at different time steps depending on the activity itself. Finally a temporal attention mechanism learns how to fuse LSTM features over time. We evaluate the method on 3 datasets. State-of-the-art results are achieved on the largest dataset for human activity recognition, namely NTU-RGB+D, as well as on the SBU Kinect Interaction dataset. Performance close to state-of-the-art is achieved on the smaller MSR Daily Activity 3D dataset. | Recurrent architectures for action recognition Most recent activity recognition methods are based on recurrent neural networks in some form. In the variant Long Short-Term Memory (LSTM) @cite_19 , a gating mechanism over an internal memory cell learns long-term and short-term dependencies in the sequential input data. Part-aware LSTMs @cite_14 separate the memory cell into part-based sub-cells and let the network learn long-term representations individually for each part, fusing the parts for output. Similarly, Du @cite_8 use bi-directional LSTM layers which fit anatomical hierarchy. Skeletons are split into anatomically-relevant parts (legs, arms, torso, etc ), so that each subnetwork in the first layers gets specialized on one part. Features are progressively merged as they pass through layers. | {
"cite_N": [
"@cite_19",
"@cite_14",
"@cite_8"
],
"mid": [
"",
"2342311830",
"1950788856"
],
"abstract": [
"",
"Recent approaches in depth-based human activity analysis achieved outstanding performance and proved the effectiveness of 3D representation for classification of action classes. Currently available depth-based and RGB+D-based action recognition benchmarks have a number of limitations, including the lack of training samples, distinct class labels, camera views and variety of subjects. In this paper we introduce a large-scale dataset for RGB+D human action recognition with more than 56 thousand video samples and 4 million frames, collected from 40 distinct subjects. Our dataset contains 60 different action classes including daily, mutual, and health-related actions. In addition, we propose a new recurrent neural network structure to model the long-term temporal correlation of the features for each body part, and utilize them for better action classification. Experimental results show the advantages of applying deep learning methods over state-of-the-art hand-crafted features on the suggested cross-subject and cross-view evaluation criteria for our dataset. The introduction of this large scale dataset will enable the community to apply, develop and adapt various data-hungry learning techniques for the task of depth-based and RGB+D-based human activity analysis.",
"Human actions can be represented by the trajectories of skeleton joints. Traditional methods generally model the spatial structure and temporal dynamics of human skeleton with hand-crafted features and recognize human actions by well-designed classifiers. In this paper, considering that recurrent neural network (RNN) can model the long-term contextual information of temporal sequences well, we propose an end-to-end hierarchical RNN for skeleton based action recognition. Instead of taking the whole skeleton as the input, we divide the human skeleton into five parts according to human physical structure, and then separately feed them to five subnets. As the number of layers increases, the representations extracted by the subnets are hierarchically fused to be the inputs of higher layers. The final representations of the skeleton sequences are fed into a single-layer perceptron, and the temporally accumulated output of the perceptron is the final decision. We compare with five other deep RNN architectures derived from our model to verify the effectiveness of the proposed network, and also compare with several other methods on three publicly available datasets. Experimental results demonstrate that our model achieves the state-of-the-art performance with high computational efficiency."
]
} |
1703.10106 | 2604906478 | We address human action recognition from multi-modal video data involving articulated pose and RGB frames and propose a two-stream approach. The pose stream is processed with a convolutional model taking as input a 3D tensor holding data from a sub-sequence. A specific joint ordering, which respects the topology of the human body, ensures that different convolutional layers correspond to meaningful levels of abstraction. The raw RGB stream is handled by a spatio-temporal soft-attention mechanism conditioned on features from the pose network. An LSTM network receives input from a set of image locations at each instant. A trainable glimpse sensor extracts features on a set of predefined locations specified by the pose stream, namely the 4 hands of the two people involved in the activity. Appearance features give important cues on hand motion and on objects held in each hand. We show that it is of high interest to shift the attention to different hands at different time steps depending on the activity itself. Finally a temporal attention mechanism learns how to fuse LSTM features over time. We evaluate the method on 3 datasets. State-of-the-art results are achieved on the largest dataset for human activity recognition, namely NTU-RGB+D, as well as on the SBU Kinect Interaction dataset. Performance close to state-of-the-art is achieved on the smaller MSR Daily Activity 3D dataset. | Attention mechanisms Human perception focuses selectively on parts of the scene to acquire information at specific places and times. In machine learning, this kind of processes is referred to as attention mechanism, and has drawn increasing interest when dealing with languages, images and other data. Integrating attention can potentially lead to improved overall accuracy, as the system can focus on parts of the data, which are most relevant to the task. In computer vision, visual attention mechanisms date as far back as the work of Itti for object detection @cite_32 . Early models were highly related to saliency maps, i.e. pixelwise weighting of image parts that locally stand out, no learning was involved. Larochelle and Hinton @cite_16 pioneered the incorporation of attention into a learning architecture by coupling Restricted Boltzmann Machines with a foveal representation. | {
"cite_N": [
"@cite_16",
"@cite_32"
],
"mid": [
"2141399712",
"2128272608"
],
"abstract": [
"We describe a model based on a Boltzmann machine with third-order connections that can learn how to accumulate information about a shape over several fixations. The model uses a retina that only has enough high resolution pixels to cover a small area of the image, so it must decide on a sequence of fixations and it must combine the \"glimpse\" at each fixation with the location of the fixation before integrating the information with information from other glimpses of the same object. We evaluate this model on a synthetic dataset and two image classification datasets, showing that it can perform at least as well as a model trained on whole images.",
"A visual attention system, inspired by the behavior and the neuronal architecture of the early primate visual system, is presented. Multiscale image features are combined into a single topographical saliency map. A dynamical neural network then selects attended locations in order of decreasing saliency. The system breaks down the complex problem of scene understanding by rapidly selecting, in a computationally efficient manner, conspicuous locations to be analyzed in detail."
]
} |
1703.10106 | 2604906478 | We address human action recognition from multi-modal video data involving articulated pose and RGB frames and propose a two-stream approach. The pose stream is processed with a convolutional model taking as input a 3D tensor holding data from a sub-sequence. A specific joint ordering, which respects the topology of the human body, ensures that different convolutional layers correspond to meaningful levels of abstraction. The raw RGB stream is handled by a spatio-temporal soft-attention mechanism conditioned on features from the pose network. An LSTM network receives input from a set of image locations at each instant. A trainable glimpse sensor extracts features on a set of predefined locations specified by the pose stream, namely the 4 hands of the two people involved in the activity. Appearance features give important cues on hand motion and on objects held in each hand. We show that it is of high interest to shift the attention to different hands at different time steps depending on the activity itself. Finally a temporal attention mechanism learns how to fuse LSTM features over time. We evaluate the method on 3 datasets. State-of-the-art results are achieved on the largest dataset for human activity recognition, namely NTU-RGB+D, as well as on the SBU Kinect Interaction dataset. Performance close to state-of-the-art is achieved on the smaller MSR Daily Activity 3D dataset. | More recently, attention mechanisms were gradually categorized into two classes. takes hard decisions when choosing parts of the input data. This leads to stochastic algorithms, which cannot be easily learned through gradient descent and back-propagation. In a seminal paper, Mnih @cite_37 proposed visual hard-attention for image classification built around a recurrent network, which implements the policy of a virtual agent. A reinforcement learning problem is thus solved during learning @cite_18 . The model selects the next location to focus on, based on past information. @cite_5 improved the approach to tackle multiple object recognition. In @cite_26 , a hard attention model generates saliency maps. Yeung @cite_46 use hard-attention for action detection with a model, which decides both which frame to observe next as well as when to emit an action prediction. | {
"cite_N": [
"@cite_37",
"@cite_26",
"@cite_18",
"@cite_5",
"@cite_46"
],
"mid": [
"2951527505",
"2342171291",
"2119717200",
"1484210532",
""
],
"abstract": [
"Applying convolutional neural networks to large images is computationally expensive because the amount of computation scales linearly with the number of image pixels. We present a novel recurrent neural network model that is capable of extracting information from an image or video by adaptively selecting a sequence of regions or locations and only processing the selected regions at high resolution. Like convolutional neural networks, the proposed model has a degree of translation invariance built-in, but the amount of computation it performs can be controlled independently of the input image size. While the model is non-differentiable, it can be trained using reinforcement learning methods to learn task-specific policies. We evaluate our model on several image classification tasks, where it significantly outperforms a convolutional neural network baseline on cluttered images, and on a dynamic visual control problem, where it learns to track a simple object without an explicit training signal for doing so.",
"Convolutional-deconvolution networks can be adopted to perform end-to-end saliency detection. But, they do not work well with objects of multiple scales. To overcome such a limitation, in this work, we propose a recurrent attentional convolutional-deconvolution network (RACDNN). Using spatial transformer and recurrent network units, RACDNN is able to iteratively attend to selected image sub-regions to perform saliency refinement progressively. Besides tackling the scale problem, RACDNN can also learn context-aware features from past iterations to enhance saliency refinement in future iterations. Experiments on several challenging saliency detection datasets validate the effectiveness of RACDNN, and show that RACDNN outperforms state-of-the-art saliency detection methods.",
"This article presents a general class of associative reinforcement learning algorithms for connectionist networks containing stochastic units. These algorithms, called REINFORCE algorithms, are shown to make weight adjustments in a direction that lies along the gradient of expected reinforcement in both immediate-reinforcement tasks and certain limited forms of delayed-reinforcement tasks, and they do this without explicitly computing gradient estimates or even storing information from which such estimates could be computed. Specific examples of such algorithms are presented, some of which bear a close relationship to certain existing algorithms while others are novel but potentially interesting in their own right. Also given are results that show how such algorithms can be naturally integrated with backpropagation. We close with a brief discussion of a number of additional issues surrounding the use of such algorithms, including what is known about their limiting behaviors as well as further considerations that might be used to help develop similar but potentially more powerful reinforcement learning algorithms.",
"We present an attention-based model for recognizing multiple objects in images. The proposed model is a deep recurrent neural network trained with reinforcement learning to attend to the most relevant regions of the input image. We show that the model learns to both localize and recognize multiple objects despite being given only class labels during training. We evaluate the model on the challenging task of transcribing house number sequences from Google Street View images and show that it is both more accurate than the state-of-the-art convolutional networks and uses fewer parameters and less computation.",
""
]
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.