instruction stringclasses 95
values | input stringlengths 141 265k | output stringlengths 15 1.61k |
|---|---|---|
Generate 2 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
Simpler Diffusion: 1.5 FID on ImageNet512 with pixel-space diffusion
## Abstract
Latent diffusion models have become the popular choice for scaling up diffusion models for high resolution image synthesis. Compared to pixel-space models that are trained end-to-end, latent models are pe... | • Compares FID scores on MSCOCO dataset
• SiD2 achieves 8.1 FID with bias=-3 and 6.7 FID with 16-step distilled version |
Generate 4 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
TWO EFFECTS, ONE TRIGGER: ON THE MODALITY GAP, OBJECT BIAS, AND INFORMATION IMBALANCE IN CONTRASTIVE VISION-LANGUAGE MODELS
## Abstract
Contrastive vision-language models (VLMs), like CLIP, have gained popularity for their versatile applicability to various downstream tasks. Despite t... | • How to measure modality gap: RMG formula based on pairwise distances
• Only a few embedding dimensions drive the modality gap
• Removal of those dimensions doesn't improve performance
• Other factors (like model size) are more important to performance |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
Learning Trimodal Relation for Audio-Visual Question Answering with Missing Modality
## Method
Fig. 2 shows the overall architecture of the proposed AVQA framework addressing missing modalities during inference. The visual modal input, audio input, and question pass through their corr... | • Audio-Visual Relation-aware Diffusion Model: Enhances feature representation of the missed modality.
• Diffusion process defined as a Gaussian noise removal process over timesteps.
• Uses weight sharing and AVQA backbone for final prediction. |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
SocraticLM: Exploring Socratic Personalized Teaching with Large Language Models
## Abstract
Large language models (LLMs) are considered a crucial technology for advancing intelligent education since they exhibit the potential for an in-depth understanding of teaching scenarios and pro... | • Proposes five metrics to evaluate LLM teaching quality: Overall Quality (Overall), Incorrect Answer Recognition Accuracy (IAR4), Correct Answer Recognition Accuracy (CAR4), Successful Explanation Rate (SER), and Successful Rejection Rate (SRR).
• Motivation: previous metrics calculate similarity between model-generat... |
Generate 3 bullet points for the "Background / Related Work" section of an academic poster. | ## Paper Content
# Paper Title
Towards Optimizing Large-Scale Multi-Graph Matching in Bioimaging
## Abstract
Multi-graph matching is an important problem in computer vision. Our task comes from bioimaging, where a set of 100 3D-microscopic images of worms have to be brought into correspondence. Surprisingly, virtua... | • Theoretical lower bound: incomplete multi-graph matching is O(n·d)
• Complete multi-graph matching is O(n·d²)
• Solvers must handle incomplete problems directly to be efficient |
Generate 1 bullet points for the "Other Content" section of an academic poster. | ## Paper Content
# Paper Title
Attention Calibration for Disentangled Text-to-Image Personalization
## Abstract
Recent thrilling progress in large-scale text-to-image (T2I) models has unlocked unprecedented synthesis quality of AI-generated content (AIGC) including image generation, 3D and video composition. Furthe... | • Datasets and code are publicly available at: https://github.com/Monalissaa/DisenDiff |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
Tiny Models are the Computational Saver for Large Models
## Method
In EE-based methods, whether to exit after a specific layer or not can be controlled by many factors. Confidence-based exiting is one of the easiest to implement. Here, EEs are designed to fulfil two primary functions:... | • By reducing the exiting threshold t, we can find how many samples S can take without hurting overall performance.
• r_match = max_t { # { m : max_i(y_S^(m,i)) ≥ t } / M } subject to correct classification.
• Reduction of complexity ΔC = E[C_TS] - C_B = (C_S / C_B) - r_match. |
Generate 4 bullet points for the "Research Motivation / Problem Background" section of an academic poster. | ## Paper Content
# Paper Title
Bones Can't Be Triangles: Accurate and Efficient Vertebrae Keypoint Estimation through Collaborative Error Revision
## Introduction
Accurate vertebrae keypoint estimation from X-ray images is crucial for effective medical diagnosis and treatment planning [20, 23], with errors having s... | • Objective: Accurate and efficient vertebrae keypoint estimation in X-ray images with minimal user intervention.
• Challenge: Similarity between vertebrae often leads to inevitable errors.
• Manual corrections are time-consuming and labor-intensive.
• Interactive keypoint estimation relies on user feedback to refine e... |
Generate 3 bullet points for the "Conclusion / Future Work" section of an academic poster. | ## Paper Content
# Paper Title
FedAWA: Adaptive Optimization of Aggregation Weights in Federated Learning Using Client Vectors
## Abstract
Federated Learning (FL) has emerged as a promising framework for distributed machine learning, enabling collaborative model training without sharing local data, thereby preservi... | • Client vectors in federated learning effectively capture relevant information about local datasets.
• The aggregated global vector is more closely aligned with the ideal update direction in federated learning (the direction that would be obtained under centralized training).
• FedAWA adaptively adjusts aggregation we... |
Generate 3 bullet points for the "Qualitative Results / Visualization" section of an academic poster. | ## Paper Content
# Paper Title
Bi-directional Contextual Attention for 3D Dense Captioning
## Experiments
Datasets. We focus on 3D dense captioning, leveraging two benchmark datasets: ScanRefer [6] and $\mathrm{Nr3D}$ [1]. These datasets offer an extensive human-generated description of 3D scenes and objects. Sca... | • Shows qualitative examples of generated captions for different methods (Scan2Cap, 3DVG-Transformer, VideoCap3D, BiCA) on ScanRefer.
• BiCA generates more accurate and detailed captions, especially for spatial relationships and object attributes.
• Failures due to low IoU are highlighted in red boxes. |
Generate 4 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
Query Efficient Black-Box Visual Prompting with Subspace Learning
## Abstract
Visual Prompt Learning (VPL) has emerged as a powerful strategy for harnessing the capabilities of large-scale pretrained models (PTMs) to tackle specific downstream tasks. However, the opaque nature of PTMs... | • The objective is to minimize downstream task loss by adapting frozen PTM via input space prompt optimization.
• Inspired by low-dimensional intrinsic subspace methods, we propose a subspace learning based black-box visual prompt approach.
• Learn projection matrices Âp and Âw by iteratively generating columns and con... |
Generate 3 bullet points for the "Qualitative Results / Visualization" section of an academic poster. | ## Paper Content
# Paper Title
Communication-Efficient Collaborative Perception via Information Filling with Codebook
## Abstract
Collaborative perception empowers each agent to improve its perceptual ability through the exchange of perceptual messages with other agents. It inherently results in a fundamental trade... | • Visualizes ego detections, ego features, information scores, and collaborator scores.
• Shows selection matrices and filled confidence maps.
• Demonstrates collaborative detections with bounding boxes overlaid on point clouds. |
Generate 3 bullet points for the "Other Content" section of an academic poster. | ## Paper Content
# Paper Title
Hybrid Proposal Refiner: Revisiting DETR Series from the Faster R-CNN Perspective
## Abstract
With the transformative impact of the Transformer, DETR pioneered the application of the encoder-decoder architecture to object detection. A collection of follow-up research, e.g., Deformable... | • Contact email: jzha0100@uni.sydney.edu.au
• Please consider citing our work if you find it helpful.
• QR codes provided for code and paper access. |
Generate 3 bullet points for the "Research Motivation / Problem Background" section of an academic poster. | ## Paper Content
# Paper Title
Leveraging Frame Affinity for sRGB-to-RAW Video De-rendering
## Abstract
Unprocessed RAW video has shown distinct advantages over sRGB video in video editing and computer vision tasks. However, capturing RAW video is challenging due to limitations in bandwidth and storage. Various met... | • Unprocessed RAW video has advantages over sRGB in editing and vision tasks.
• Capturing RAW video is challenging due to bandwidth and storage limitations.
• Previous sRGB-to-RAW methods store per-image metadata; this work uses frame affinity instead. |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
NVComposer: Boosting Generative Novel View Synthesis with Multiple Sparse and Unposed Images
## Abstract
Recent advancements in generative models have significantly improved novel view synthesis (NVS) from multi-view data. However, existing methods depend on external multiview alignme... | • Proposes a dual-stream diffusion model that generates image-pose bundles jointly, based on video prior.
• Introduces a geometry-aware feature alignment adapter that distills geometric prior during training.
• Architecture includes CLIP encoder, latent encoder/decoder, and dual streams for image and pose generation. |
Generate 4 bullet points for the "Research Motivation / Problem Background" section of an academic poster. | ## Paper Content
# Paper Title
Virtual Immunohistochemistry Staining for Histological Images Assisted by Weakly-supervised Learning
## Abstract
Recently, virtual staining technology has greatly promoted the advancement of histopathology. Despite the practical successes achieved, the outstanding performance of most ... | • H&E staining is common but lacks cellular detail needed for pathology.
• IHC staining provides necessary detail but is time-consuming and labor-intensive.
• Virtual staining via deep learning is promising but often requires hard-to-obtain paired images.
• We propose confusion-GAN, a weakly-supervised method that does... |
Generate 3 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
ImViD: Immersive Volumetric Videos for Enhanced VR Engagement
## Abstract
User engagement is greatly enhanced by fully immersive multi-modal experiences that combine visual and auditory stimuli. Consequently, the next frontier in VR/AR technologies lies in immersive volumetric videos ... | • Benchmark established for SOTA 4DGS-based reconstruction methods across three paradigms: 4DGS, 4DRotor, SpacetimeGS.
• Quantitative and qualitative comparison between STG++ and baselines on multiple scenes.
• Metrics include PSNR, SSIM, LPIPS for different scene types and capture strategies. |
Generate 1 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
Movie Weaver: Tuning-Free Multi-Concept Video Personalization with Anchored Prompts
## Abstract
Video personalization, which generates customized
videos using reference images, has gained significant attention. However, prior methods typically focus on single-concept personalization,... | • For each video-text pair: (1) Llama-3 generates anchor prompts, (2) body masks are extracted, (3) Concepts are linked to images via CLIP. |
Generate 3 bullet points for the "Other Content" section of an academic poster. | ## Paper Content
# Paper Title
CASAGPT: Cuboid Arrangement and Scene Assembly for Interior Design
## Abstract
We present a novel approach for indoor scene synthesis, which learns to arrange decomposed cuboid primitives to represent 3D objects within a scene. Unlike conventional methods that use bounding boxes to de... | • Code: https://github.com/CASAGPT/CASA-GPT
• Dataset: https://github.com/CASAGPT/3DFRONT-NC
• Contact: weitaofeng@mail.ustc.edu.cn, Ge0rg3Fe |
Generate 3 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
Understanding Video Transformers via Universal Concept Discovery
## Abstract
This paper studies the problem of concept-based interpretability of transformer representations for videos. Concretely, we seek to explain the decision-making process of video transformers based on high-level... | • Validates VTCD with concept attribution curves, outperforming occlusion and gradient-based methods.
• Tubelet proposals + CRIS are superior to baselines in matching groundtruth masks.
• Plots show positive and negative perturbation effects on mIoU and Accuracy. |
Generate 2 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
## Abstract
We introduce X-Ray, a novel 3D sequential representation inspired by the penetrability of x-ray scans. X-Ray transforms a 3D object into a series of surface frames at different layers, making it suitable for generating 3D models from images. Our method utilizes ray casting from the camera... | • Applies ray casting algorithm to encode a 3D object mesh into the proposed X-Ray representation.
• Formula: X_ij = X_r = (H_r, D_r, N_r, C_r) ∈ ℝ^L×8. |
Generate 3 bullet points for the "Method Overview / Framework" section of an academic poster. | ## Paper Content
# Paper Title
SCSegamba: Lightweight Structure-Aware Vision Mamba for Crack Segmentation in Structures
## Abstract
Pixel-level segmentation of structural cracks across various scenarios remains a considerable challenge. Current methods encounter challenges in effectively modeling crack morphology a... | • We propose a novel lightweight vision Mamba network, SCSegamba, for crack segmentation that captures morphological and irregular texture cues with low computational resources.
• We design the SAVSS with a lightweight GBC Convolution and a SASS scanning strategy to enhance handling of irregular texture cues.
• We eval... |
Generate 3 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
EAP-GS: Efficient Augmentation of Pointcloud for 3D Gaussian Splitting in Few-shot Scene Reconstruction
## Abstract
3D Gaussian splattering (3DGS) has shown impressive performance in 3D scene reconstruction. However, it suffers from severe degradation when the number of training views... | • Qualitative Comparison: EAP-GS (Ours) outperforms GT, 3DGS, DRGS, FSGS, and CoR-GS across multiple scenes (plant, bonsai, trex, garden) with sharper details and fewer artifacts.
• Pointcloud Optimization: Before/After comparison shows significant point increase (e.g., Fern from 2000 to 9292 points, Bonsai from 138 to... |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
DO LLMs RECOGNIZE YOUR PREFERENCES? EVALUATING PERSONALIZED PREFERENCE FOLLOWING IN LLMs
## Abstract
Large Language Models (LLMs) are increasingly used as chatbots, yet their ability to personalize responses to user preferences remains limited. We introduce PREFEVAL, a benchmark for e... | • Generation: LLM generates response based on user preference.
• Classification: LLM classifies whether a response follows user preference.
• LLM-as-a-Judge Eval: Automatic evaluation with ground truth choice. |
Generate 2 bullet points for the "Background / Related Work" section of an academic poster. | ## Paper Content
# Paper Title
Investigating Compositional Challenges in Vision-Language Models for Visual Grounding
## Abstract
Pre-trained vision-language models (VLMs) have achieved high performance on various downstream tasks, which have been widely used for visual grounding tasks in a weakly supervised manner.... | • Established based on Visual Genome, ARPGrounding contains 11,425 samples and evaluates the compositional understanding of VLMs in three dimensions: attribute, relation and priority.
• We test whether the model can pick the correct object according to the text. |
Generate 2 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
Not Only Text: Exploring Compositionality of Visual Representations in Vision-Language Models
## Abstract
Vision-Language Models (VLMs) learn a shared feature space for text and images, enabling the comparison of inputs of different modalities. While prior works demonstrated that VLMs... | • Unlike linear decomposition, our visual GDE are strong classifiers in compositional classification.
• This indicates they encode semantically meaningful information. |
Generate 2 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
Embracing Events and Frames with Hierarchical Feature Refinement Network for Object Detection
## Method
In this section, we first introduce preliminaries, including the working principles of the event camera and event representation. Subsequently, we delve into a detailed illustration... | • The model uses a dual-stream backbone (RGB Frame and Voxel Grid), CAFR modules, FPN, and a detection head.
• CAFR modules refine features across layers using cross-modality interaction and adaptive refinement. |
Generate 3 bullet points for the "Qualitative Results / Visualization" section of an academic poster. | ## Paper Content
# Paper Title
Learning Heterogeneous Tissues with Mixture of Experts for Gigapixel Whole Slide Images
## Abstract
Analyzing gigapixel Whole Slide Images (WSIs) is challenging due to the complex pathological tissue environment and the absence of target-driven domain knowledge. Previous methods incor... | • Shows whole slide image with expert assignment map and ROI with expert preference.
• Displays six expert visualizations with prior supervised and free experts.
• Highlights not chosen instances for each expert. |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
Siyuan Duan $^{*1}$ Wenyuan Wu $^{*1}$ Peng Hu $^{1}$ Zhenwen Ren $^{2}$ Dezhong Peng $^{13}$ Yuan Sun $^{13}$
## Abstract
Physics-informed neural networks (PINN) aim to constrain the outputs and gradients of deep learning models to satisfy specified governing physics equat... | • The method includes sample point selection, separable learning, prediction, and loss computation with sample weights.
• A cognitive training scheduler adjusts sample point weights over epochs from easy to hard.
• The weight formula combines epoch change and sample change components based on difficulty measure. |
Generate 3 bullet points for the "Conclusion / Future Work" section of an academic poster. | ## Paper Content
# Paper Title
Multi-Sentence Grounding for Long-term Instructional Video
## Conclusion
To conclude, we have established an automatic pipeline for constructing a high-quality video-text dataset for multi-sentence grounding in large-scale instructional videos. We have investigated the factors potenti... | • Dataset: HowToStep, a large-scale dataset efficiently produced by an automatic, scalable pipeline using LLMs.
• Architecture: NaSVA, a Transformer-based architecture for multi-sentence grounding.
• Benchmarks: SOTA approach on multi-sentence grounding tasks. |
Generate 3 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
Degradation-Aware Feature Perturbation for All-in-One Image Restoration
## Abstract
All-in-one image restoration aims to recover clear images from various degradation types and levels with a unified model. Nonetheless, the significant variations among degradation types present challen... | • Quantitative comparisons on three/five tasks show DFPIR outperforms state-of-the-art methods.
• On average, DFPIR achieves 0.45 dB PSNR gain over InstructIR on full RGB images.
• Results reported for PSNR (dB) and SSIM metrics across multiple degradation types and datasets. |
Generate 4 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
AUFormer: Vision Transformers are Parameter-Efficient Facial Action Unit Detectors
## Experiments
Datasets. We evaluate the proposed AUFormer on three datasets: BP4D [71] and DISFA [39] in the macro-expression domain, and CASME II [64] in the micro-expression domain. For BP4D and DISF... | • Within-domain evaluations on BP4D and DISFA show AUFormer outperforms baselines.
• Cross-domain evaluations between BP4D and DISFA demonstrate robust generalization.
• Data efficiency evaluation on BP4D shows AUFormer maintains performance with fewer training samples.
• Micro-expression domain evaluation on CASME II ... |
Generate 2 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
Continuous, Subject-Specific Attribute Control in T2I Models by Identifying Semantic Directions
## Abstract
Recent advances in text-to-image (T2I) diffusion models have significantly improved the quality of generated images. However, providing efficient control over individual subject... | • Table compares methods on Subject-Specificity, Disentanglement, and Performance.
• Our method achieves high subject-specificity (3.15) and disentanglement (0.73) with competitive performance (23.0s). |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
Scaling up Image Segmentation across Data and Tasks
## Abstract
Traditional segmentation models, while effective in isolated tasks, often fail to generalize to more complex and open-ended segmentation problems, such as free-form, open-vocabulary, and in-the-wild scenarios. To bridge t... | • High-confidence proposals from transformer encoder are selected as object queries for final predictions.
• Performance on semantic segmentation is worse compared to learnable queries.
• Learnable queries excel at large background regions; conditional queries specialize in local, instance-level features. |
Generate 2 bullet points for the "Research Motivation / Problem Background" section of an academic poster. | ## Paper Content
# Paper Title
Text-DiFuse: An Interactive Multi-Modal Image Fusion Framework based on Text-modulated Diffusion Model
## Abstract
Existing multi-modal image fusion methods fail to address the compound degradations presented in source images, resulting in fusion images plagued by noise, color bias, i... | • Current fusion methods falter in scenes with degradation, especially composite degradation.
• They overlook the specificity of foreground objects, lacking interaction support. |
Generate 2 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
ShareGPT4V: Improving Large Multi-Modal Models with Better Captions
## Method
In this section, we detail the process of developing the ShareGPT4V series. Subsection 3.2 elaborates how we created the ShareGPT4V dataset, including how we harnessed GPT4-Vision to generate 100K high-quali... | • Introduces an intermediate stage between traditional connector training and supervised fine-tuning to leverage high-quality image-caption data.
• Fine-tunes LLM, connector, and vision encoder layers using 1.2M high-quality data to maximize visual-language alignment. |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
Noise Calibration and Spatial-Frequency Interactive Network for STEM Image Enhancement
## Abstract
Scanning Transmission Electron Microscopy (STEM) enables the observation of atomic arrangements at subangstrom resolution, allowing for atomically resolved analysis of the physical and c... | • Introduce STEM image noise model: background (2-D Perlin), scan (1-D Perlin per row), and pointwise (Tukey’s lambda) noise.
• Calibrate scan and pointwise noise using parameter plots based on image gradient and pixel value statistics.
• Construct a more general synthetic dataset with periodic, defect, random, HAADF, ... |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
Suitability Filter: A Statistical Framework for Classifier Evaluation in Real-World Deployment Settings
## Abstract
Deploying machine learning models in safety-critical domains poses a key challenge: ensuring reliable model performance on downstream user data without access to ground ... | • Adjusted margin m' = m + Δtest - Δu accounts for distribution shifts.
• Δtest and Δu represent accuracy alignment differences between test and user data.
• Ensures the suitability filter remains reliable under distribution shifts. |
Generate 4 bullet points for the "Ablation Study" section of an academic poster. | ## Paper Content
## Abstract
The advancement of Large Vision Language Models (LVLMs) has significantly improved multimodal understanding, yet challenges remain in video reasoning tasks due to the scarcity of high-quality, large-scale datasets. Existing video question-answering (VideoQA) datasets often rely on costly ... | • Ablation shows GT-CoT improves performance by +38.82.
• Baseline without CoT drops performance by -0.99.
• w/o CoT and w/o Bbox show reduced accuracy.
• 2.5B model achieves 2.36 gain over baseline. |
Generate 4 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
CoGS: Controllable Gaussian Splitting
## Abstract
Capturing and re-animating the 3D structure of articulated objects present significant barriers. On one hand, methods requiring extensively calibrated multi-view setups are prohibitively complex and resource-intensive, limiting their p... | • A local rigidity loss ensures neighboring Gaussians move consistently with rigid body transform over time.
• A rotational loss maintains consistency in rotations among nearby Gaussians across time.
• With a robust dynamic method, it is extended to controllable scenarios with control signals and 3D masks in three phas... |
Generate 4 bullet points for the "Implementation Details" section of an academic poster. | ## Paper Content
# Paper Title
ViGoR: Improving Visual Grounding of Large Vision Language Models with Fine-Grained Reward Modeling
## Method
Human Preference Data Collection. We design a system to incorporate human judgment and preference — considered the most reliable ground truth signal — into our model training.... | • Dataset includes 15,440 samples with fine-grained human annotations.
• Each sample contains input image, sampled LVLM output, and annotator judgement on correctness, creativity, and detail.
• Annotations cover object count, object existence, attribute accuracy, relationship accuracy, reasoning, and detail level.
• Da... |
Generate 3 bullet points for the "Implementation Details" section of an academic poster. | ## Paper Content
# Paper Title
SocraticLM: Exploring Socratic Personalized Teaching with Large Language Models
## Abstract
Large language models (LLMs) are considered a crucial technology for advancing intelligent education since they exhibit the potential for an in-depth understanding of teaching scenarios and pro... | • Motivation: enhance diversity and robustness in DiagM to improve teaching abilities.
• Data Augmentation: sample 2000 'Teacher-Student' dialogues from DiagM and extend to 2000/6000/10000/4000 single-round dialogues.
• Aims to improve model generalization and adaptability to varied student responses. |
Generate 4 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
UniMD: Towards Unifying Moment Retrieval and Temporal Action Detection
## Experiments
Text Encoder. To establish a correspondence between the action names from TAD and the natural language descriptions from MR, we utilize the text encoder from a well-trained image-text alignment model... | • Results on Ego4D: UniMD-Sync achieves 23.25 mAP, 44.80 R1@50, 14.16 R1@30, 10.06 R1@50 for TAD; 26.95 R5@70 for MR.
• Results on ActivityNet: UniMD-Sync achieves 39.83 mAP, 60.29 R1@50, 80.54 R5@70 for TAD; 57.94 R5@70 for MR.
• Results on Charades: UniMD-Sync achieves 26.53 mAP, 63.98 R1@50, 44.46 R1@30, 91.94 R5@70... |
Generate 2 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
GenH2R: Learning Generalizable Human-to-Robot Handover via Scalable Simulation, Demonstration, and Imitation
## Abstract
This paper presents GenH2R, a framework for learning generalizable vision-based human-to-robot (H2R) handover skills. The goal is to equip robots with the ability t... | • We propose a 4D imitation learning method that factors sequential point cloud observations into geometry and motion parts by estimated flow information, facilitating policy learning by revealing the current scene state.
• The imitation objective is augmented by a forecasting objective that predicts the future motion ... |
Generate 3 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods
## Abstract
In order to gain insights about the decision-making of different visual recognition backbones, we propose two methodologies, sub-explanation counting and cross-testing, that systemati... | • Tabular comparison of MSEs and number of sub-explanations across different models.
• Shows performance differences between older CNNs, newer CNNs, and Transformers.
• Highlights the impact of distillation on model performance. |
Generate 3 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
Layered Image Vectorization via Semantic Simplification
## Abstract
This work presents a progressive image vectorization technique that reconstructs the raster image as layer-wise vectors from semantic-aligned macro structures to finer details. Our approach introduces a new image simp... | • Qualitative & Quantitative Comparison with four state-of-the-art methods: DiffVG, O&R, LIVE, SGLIVE.
• Our method achieves superior visual quality and quantitative scores (LPIPS, MSE, CLIP Score) across varying path numbers.
• Layered Representation and Editing enables easy recoloring: vectors inside ice cream ball a... |
Generate 2 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
Inlier Confidence Calibration for Point Cloud Registration
## Abstract
Inliers estimation constitutes a pivotal step in partially overlapping point cloud registration. Existing methods broadly obey coordinate-based scheme, where inlier confidence is scored through simply capturing coo... | • Constructing Geometric Constraints: Defines geometric signal buried in point cloud using point pairs and constraints.
• Byproduct for Optimization: Introduces geometric structure consistency loss combining point pair differences and constraints. |
Generate 2 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
A Bias-Free Training Paradigm for More General AI-generated Image Detection
## Abstract
Successful forensic detectors can produce excellent results in supervised learning benchmarks but struggle to transfer to real-world applications. We believe this limitation is largely due to inade... | • Objects are replaced with generated ones (same or different category)
• Besides default inpainting, which regenerates the whole image, we also consider a version with the original background restored |
Generate 4 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
ROD-MLLM: Towards More Reliable Object Detection in Multimodal Large Language Models
## Abstract
Multimodal large language models (MLLMs) have demonstrated strong language understanding and generation capabilities, excelling in visual tasks like referring and grounding. However, due t... | • Architecture: OVD for low-level localization & LLM for high-level comprehension.
• High-Level Comprehension: LLM internal inference generates region proposals.
• Low-Level Localization: OVD refines object locations using visual encoder and ROIAlign.
• Dataset: Richer object semantics & 0-to-multiple objects & confusi... |
Generate 4 bullet points for the "Research Motivation / Problem Background" section of an academic poster. | ## Paper Content
# Paper Title
SURE: SSurvey REcipes for building reliable and robust deep networks
## Abstract
In this paper, we revisit techniques for uncertainty estimation within deep neural networks and consolidate a suite of techniques to enhance their reliability. Our investigation reveals that an integrated... | • Model robustness in handling complex real-world data challenges, such as long-tailed classification, learning with noisy labels and data corruptions.
• Contribution 1: Simple and effective approach SURE for building reliable and robust deep networks.
• Contribution 2: SOTA performance in failure prediction across var... |
Generate 3 bullet points for the "Ablation Study" section of an academic poster. | ## Paper Content
# Paper Title
InfMAE: A Foundation Model in The Infrared Modality
## Experiments
As depicted in Fig. 3, the proposed InfMAE architecture consists of three principal modules: 1) the mask block generation module, 2) the multi-scale encoder module, and 3) the infrared decoder module. In the mask block... | • Ablation study on the influence of different settings.
• Decoder depth analysis showing performance at depths 2, 4, 8, 12.
• Strides setting and pre-train epochs analysis. |
Generate 3 bullet points for the "Background / Related Work" section of an academic poster. | ## Paper Content
# Paper Title
Unsupervised Salient Instance Detection
## Abstract
The significant amount of manual efforts in annotating pixel-level labels has triggered the advancement of unsupervised saliency learning. However, without supervision signals, state-of-the-art methods can only infer region-level sal... | • Manual pixel-level annotation efforts have driven unsupervised saliency learning, but without supervision, state-of-the-art methods only infer region-level saliency.
• Existing methods like FOUND and SelfMask fail to differentiate salient instances, especially in complex scenes with strong semantic instances.
• Our m... |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
Omni-Q: Omni-Directional Scene Understanding for Unsupervised Visual Grounding
## Abstract
Unsupervised visual grounding methods alleviate the issue of expensive manual annotation of image-query pairs by generating pseudo-queries. However, existing methods are prone to confusing the s... | • Object Perception Module: extracts object locations and rich descriptions.
• 3D Spatial Relation Module: assembles relative and absolute positions to locate objects.
• Spatial Graph Module: leverages graph structure to establish spatial connections between objects. |
Generate 3 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
## Abstract
Automatic program generation has long been a fundamental challenge in computer science. Recent benchmarks have shown that large language models (LLMs) can effectively generate code at the function level, make code edits, and solve algorithmic coding tasks. However, to achieve full automat... | • On average, secure solutions are 5% longer than insecure programs.
• Certain frameworks incur higher security overhead, e.g., secure JS-Express backends are >10% longer.
• A champion among scenarios, secure implementations of the calculator backend are 15% longer. |
Generate 3 bullet points for the "Ablation Study" section of an academic poster. | ## Paper Content
# Paper Title
Beta-Tuned Timestep Diffusion Model
## Experiments
In this section, we first provide an implementation details of the experimental setup in section 4.1. Then, we provide quantitative comparison of B-TTDM with other state-of-the-art methods in section 4.2. In Section 4.3, we assess the... | • Ablation study on Beta Distribution parameters α and β shows impact on FID, sFID, Recall, and Precision.
• Optimal performance achieved at α=0.80, β=1 with FID 13.81 and sFID 45.25.
• Lower α values (0.70) degrade performance, indicating sensitivity to parameter choice. |
Generate 2 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
Spiking Transformer with Spatial-Temporal Attention
## Abstract
Spike-based Transformer presents a compelling and energy-efficient alternative to traditional Artificial Neural Network (ANN)-based Transformers, achieving impressive results through sparse binary computations. However, e... | • STAtten achieves comparable performance to larger models with fewer parameters.
• Spatial-temporal processing incurs no additional energy consumption compared to spatial-only attention. |
Generate 3 bullet points for the "Method Overview / Framework" section of an academic poster. | ## Paper Content
## Introduction
The core objective for video understanding lies in mastering spatiotemporal representations, which presents two formidable challenges: large spatiotemporal redundancy in short video clips and complex spatiotemporal dependencies in long contexts. Although the once-dominant 3D CNNs [9,2... | • Introduces the core question: adapting Mamba architecture for video understanding.
• Highlights the proposed VideoMamba framework with bidirectional scanning.
• Emphasizes efficiency and scalability as key advantages. |
Generate 2 bullet points for the "Background / Related Work" section of an academic poster. | ## Paper Content
# Paper Title
SnowMaster: Comprehensive Real-world Image Desnowing via MLLM with Multi-Model Feedback Optimization
## Abstract
Snowfall presents significant challenges for visual data processing, necessitating specialized desnowing algorithms. However, existing models often fail to generalize effec... | • We propose RealSnow10K, the largest real-world snowfall image dataset, containing over 10K high-quality snowfall images.
• We utilize MLLM with Multiple Choice Questions for data cleaning and annotation, where higher scores indicate greater snowfall intensity. |
Generate 2 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
MELFUSION: Synthesizing Music from Image and Language Cues using Diffusion Models
## Abstract
Music is a universal language that can communicate emotions and feelings. It forms an essential part of the whole spectrum of creative media, ranging from movies to social media posts. Machin... | • MeLFusion offers significant gains over state-of-the-art text-to-music methods (upper half) and adapted baselines (lower half) across multiple metrics.
• Performance varies with prompt length and verboisty. |
Generate 3 bullet points for the "Background / Related Work" section of an academic poster. | ## Paper Content
# Paper Title
Producing and Leveraging Online Map Uncertainty in Trajectory Prediction
## Abstract
High-definition (HD) maps have played an integral role in the development of modern autonomous vehicle (AV) stacks, albeit with high associated labeling and maintenance costs. As a result, many recent... | • Dataset: nuScenes
• Mapping Models: MapTR series and StreamMapNet
• Prediction Models: DenseTNT (GNN) and HiVT (Transformer) |
Generate 3 bullet points for the "Qualitative Results / Visualization" section of an academic poster. | ## Paper Content
# Paper Title
DynPose: Largely Improving the Efficiency of Human Pose Estimation by a Simple Dynamic Framework
## Abstract
Top-down approaches for human pose estimation (HPE) have reached a high level of sophistication, exemplified by models such as HRNet and ViTpose. Nonetheless, the low efficienc... | • Shows visualization samples with different scores: Easy, Hard, Extremely Hard.
• Orange and green annotations represent predictions of ResNet-50 and ViT-L, respectively.
• Illustrates pose sample diversity and model performance variation across difficulty levels. |
Generate 4 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
TUMTraf V2X Cooperative Perception Dataset
## Abstract
Cooperative perception offers several benefits for enhancing the capabilities of autonomous vehicles and improving road safety. Using roadside sensors in addition to onboard sensors increases reliability and extends the sensor ran... | • Roadside Camera-LiDAR Fusion
• Cooperative Fusion via Vehicle-Infrastructure Max Fuser (PillarGrid)
• Vehicle Camera-LiDAR Fusion
• Collaborative 3D Detections with 3D Detection Head |
Generate 1 bullet points for the "Other Content" section of an academic poster. | ## Paper Content
# Paper Title
Decompose-and-Compose: A Compositional Approach to Mitigating Spurious Correlation
## Abstract
While standard Empirical Risk Minimization (ERM) training is proven effective for image classification on in-distribution data, it fails to perform well on out-of-distribution samples. One o... | • List of 7 academic references related to spurious correlation mitigation, group robustness, and contrastive learning. |
Generate 6 bullet points for the "Research Motivation / Problem Background" section of an academic poster. | ## Paper Content
# Paper Title
Adversarial Backdoor Attack by Naturalistic Data Poisoning on Trajectory Prediction in Autonomous Driving
## Abstract
In autonomous driving, behavior prediction is fundamental for safe motion planning, hence the security and robustness of prediction models against adversarial attacks ... | • Security of DNNs for safety-critical AD systems is vital
• No existing backdoor attack methods on trajectory prediction
• An effective train-time attack should be:
• Unnoticeable: Backdoor-injected model should behave similarly to clean model on original data
• Effective: Poisoned samples should cause malicious behav... |
Generate 4 bullet points for the "Background / Related Work" section of an academic poster. | ## Paper Content
# Paper Title
MobileNetV4: Universal Models for the Mobile Ecosystem
## Introduction
Efficient on-device neural networks not only enable fast, real-time and interactive experiences, but also avoid streaming private data through the public internet. However, the computational constraints of mobile d... | • Problems: MAC counting is poorly correlated with latency; Measuring latency on real hardware is difficult; Hardware performance is hard to generalize.
• Our Solution: Use the Roofline Model as an abstract hardware performance model; Characterize hardware behavior by PeakMACs & PeakMBW; Layer latency is determined by ... |
Generate 2 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
TSAM: Temporal SAM Augmented with Multimodal Prompts for Referring Audio-Visual Segmentation
## Abstract
Referring audio-visual segmentation (Ref-AVS) aims to segment objects within audio-visual scenes using multimodal cues embedded in text expressions. While the Segment Anything Mode... | • Performance comparison table showing TSAM outperforming baselines on Ref-AVS benchmarks.
• Ablation study table demonstrating contribution of each component (TB, TMFL, DPM, SPM, CM, AM) to final performance. |
Generate 3 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
Task-aligned Part-aware Panoptic Segmentation through Joint Object-Part Representations
## Abstract
Part-aware panoptic segmentation (PPS) requires (a) that each foreground object and background region in an image is segmented and classified, and (b) that all parts within foreground o... | • TAPPS outperforms strong baselines on Pascal and Cityscapes datasets.
• Compared to existing SOTA (PPF++, VIRReq), TAPPS achieves higher scores across PartPQ, PartSQ, and PQ metrics.
• Results show consistent gains in both object and part segmentation quality. |
Generate 3 bullet points for the "Background / Related Work" section of an academic poster. | ## Paper Content
# Paper Title
HoloADMM: High-Quality Holographic Complex Field Recovery
## Introduction
Dennis Gabor introduced holography in his seminal work [13] where he demonstrated true three dimensional imaging capability by simultaneously recording coherent light's intensity and direction, i.e., phase, with... | • Explains holographic concepts: coherent light, interference, and phase encoding.
• Describes DIHM prototype: lens-free digital in-line holographic microscope.
• Shows system components: laser illumination, object plane, detector, and piezo motor stage. |
Generate 3 bullet points for the "Implementation Details" section of an academic poster. | ## Paper Content
# Paper Title
Collaborative Control for Geometry-Conditioned PBR Image Generation
## Method
We wish to train a PBR diffusion model $\mathcal{D}_{pbr}$ that models the reverse denoising process for PBR images as represented in the latent space of a VAE [48], representing the data distribution $p(... | • The frozen base model is controlled to output render-like RGB images.
• Fixing the environment to be camera-aligned simplifies the control problem.
• This yields higher quality outputs by reducing pose and lighting variability. |
Generate 2 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
SeiT++: Masked Token Modeling Improves Storage-efficient Training
## Experiments
In this section, we validate the effectiveness of SeiT++ on various scenarios. First, we evaluate the performance of our method in the context of storage-efficient ImageNet-1k classification and transfer ... | • Loss curves show SeiT++ converges faster and achieves lower loss than SeiT in MTM pre-training and ImageNet-1k classification.
• Test loss curves confirm SeiT++ generalizes better on unseen data. |
Generate 3 bullet points for the "Method Overview / Framework" section of an academic poster. | ## Paper Content
# Paper Title
GALA: Generating Animatable Layered Assets from a Single Scan
## Abstract
We present GALA, a framework that takes as input a single-layer clothed 3D human mesh and decomposes it into complete multi-layered 3D assets. The outputs can then be combined with other assets to create novel c... | • We can decompose any asset from a scan by lifting 2D segmentations into 3D, and completing the missing parts via 2D diffusion priors.
• Pose-guided SDS loss successfully eliminates the undesired artifacts of a single-scan canonicalization process.
• Through a series of decomposition and completion, we can achieve ani... |
Generate 4 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
OP-Align: Object-level and Part-level Alignment for Self-supervised Category-level Articulated Object Pose Estimation
## Experiments
Datasets: We use a synthetic dataset generated by authors of EAP [24] and our real-world dataset for evaluation. The synthetic dataset contains laptop, ... | • Partially observed point cloud generated with random camera position
• Compared with other supervised and self-supervised methods
• Achieved state-of-the-art performance and real-time inference speed
• Mean Error Results table shows Ours outperforming baselines in most metrics |
Generate 2 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
NGP-RT: Fusing Multi-Level Hash Features with Lightweight Attention for Real-Time Novel View Synthesis
## Task
Section Title: Method
Generate poster content for this section. | • Lightweight Attention: Adaptively prioritizes explicit multi-level hash features using a lightweight attention mechanism; attention weights vary spatially for texture and geometry features.
• Occupancy Distance Grid: Stores distance to nearest occupied voxel to reduce global memory access during ray marching. |
Generate 2 bullet points for the "Ablation Study" section of an academic poster. | ## Paper Content
# Paper Title
DENSE VIDEO OBJECT CAPTIONING FROM DISJOINT SUPERVISION
## Abstract
We propose a new task and model for dense video object captioning - detecting, tracking and captioning trajectories of objects in a video. This task unifies spatial and temporal localization in video, whilst also requ... | • We compare different backbone models (ResNet50, ViT-B) and training strategies (pretrain, finetune, zero-shot).
• Our method with ViT-B backbone achieves 61.9 CHOTA in finetuned setting and 54.1 in zero-shot, outperforming baselines. |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
Removing Reflections from RAW Photos
## Abstract
We describe a system to remove real-world reflections from images for consumer photography. Our system operates on linear (RAW) photos, and accepts an optional contextual photo looking in the opposite direction (e.g. the "selfie" camera... | • Shows pipeline from incident light fields to rendered image
• Highlights that photometric principles are violated in standard pipeline
• Notes that light isn't linearly related to pixels, illuminant colors are mixed, and illuminant powers are removed |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
DreamStruct: Understanding Slides and User Interfaces via Synthetic Data Generation
## Method
Instead of relying on traditional methods for building visual understanding datasets, which require human annotation of inputs (e.g., screenshots), DreamStruct begins with an abstract specifi... | • Generates design concepts from seed examples, guidelines, and semantic labels using a description generation function.
• Condition: BERTScore of generated concept must be below 0.7.
• Includes generated slide description for a Japanese verb conjugation slide with bar chart and transformation table. |
Generate 2 bullet points for the "Conclusion / Future Work" section of an academic poster. | ## Paper Content
# Paper Title
DirectTriGS: Triplane-based Gaussian Splitting Field Representation for 3D Generation
## Abstract
We present DirectTriGS, a novel framework designed for 3D object generation with Gaussian Splatting (GS). GS-based rendering for 3D content has gained considerable attention recently. How... | • DirectTriGS mainly consists of 3 parts: 1) a light-weight triplane representation for 3D object with the format of Gaussian Splatting, 2) a fully differentiable TriRenderer which can decode triplane to original GS and render it to multi-view images, and 3) the triplane VAE and staged diffusion model for the whole gen... |
Generate 3 bullet points for the "Research Motivation / Problem Background" section of an academic poster. | ## Paper Content
# Paper Title
Holodepth: Programmable Depth-Varying Projection via Computer-Generated Holography
## Introduction
Projectors have found widespread use in the real world today. They are ubiquitous in entertainment and education, used to show movies and messages on large screens for big audiences [6].... | • Traditional projectors form a single desired image at a single plane.
• Content at any other plane is a blurred version of the same image.
• Our goal is to build a projector that can simultaneously form different images at different distances. |
Generate 3 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
BEYOND SQUARED ERROR: EXPLORING LOSS DESIGN FOR ENHANCED TRAINING OF GENERATIVE FLOW NETWORKS
## Abstract
Generative Flow Networks (GFlowNets) are a novel class of generative models designed to sample from unnormalized distributions and have found applications in various important tas... | • Reports results on bit-sequence generation task using TB, DB, and STB algorithms with different loss functions.
• Table 3 shows the number of runs finding all modes within 250k steps and median steps to convergence.
• Table 4 reports Spearman correlation between predicted and true rewards, indicating how well modal c... |
Generate 3 bullet points for the "Qualitative Results / Visualization" section of an academic poster. | ## Paper Content
# Paper Title
Mitigating the Human-Robot Domain Discrepancy in Visual Pre-training for Robotic Manipulation
## Abstract
Learning generalizable visual representations across different embodied environments is essential for effective robotic manipulation in real-world scenarios. However, the limited ... | • Shows robot performing five manipulation tasks: putting fruit in plate, stacking cups, etc.
• Demonstrates successful execution of tasks using the aligned representation.
• Visual examples validate real-world applicability of the method. |
Generate 3 bullet points for the "Conclusion / Future Work" section of an academic poster. | ## Paper Content
# Paper Title
ProMark: Proactive Diffusion Watermarking for Causal Attribution
## Abstract
Generative AI (GenAI) is transforming creative workflows through the capability to synthesize and manipulate images via high-level prompts. Yet creatives are not well supported to receive recognition or rewar... | • First proactive scheme for concept attribution in generative models.
• Proactive attribution outperforms passive methods in accuracy.
• Supports single-concept and multi-concept attribution with flexible secret embedding. |
Generate 2 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
Ground-V: Teaching VLMs to Ground Complex Instructions in Pixels
## Abstract
This work presents a simple yet effective workflow for automatically scaling instruction-following data to elicit pixel-level grounding capabilities of VLMs under complex instructions. In particular, we addre... | • Table 3 compares performance of GSVA, GSVA-G5, LISA, LISA-G5, PSALM, and PSALM-G5 across five challenging real-world referring segmentation scenarios.
• PSALM-G5 achieves top scores in most categories, e.g., 58.7 gIoU on Abstract, 54.7 on FG, 39.5 on Multi-object, 33.9 N-Acc on Hallucination, 72.9 gIoU on Reasoning, ... |
Generate 4 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
BACON: Improving Clarity of Image Captions via Bag-of-Concept Graphs
## Abstract
Advancements in large Vision-Language Models have brought precise, accurate image captioning, vital for advancing multi-modal image understanding and processing. Yet these captions often carry lengthy, in... | • Convert BACON-style caption into VLM-readable string format.
• Employ ICL to teach VLMs to output in specified format.
• Use grounding DINO to obtain required bounding boxes.
• BACON-style caption includes object list, scene description, and global description. |
Generate 2 bullet points for the "Qualitative Results / Visualization" section of an academic poster. | ## Paper Content
# Paper Title
A Semi-supervised Nighttime Dehazing Baseline with Spatial-Frequency Aware and Realistic Brightness Constraint
## Abstract
Existing research based on deep learning has extensively explored the problem of daytime image dehazing. However, few studies have considered the characteristics ... | • Figure 5 presents visual results on real-world nighttime haze (RWNH) dataset, comparing 10 methods including Hazy, MRP, OSFD, GD, MSBDN, 4KDehazing, AECRNet, DeHamer, DF, MITNet, Fourier, and Ours.
• Our method preserves realistic brightness and suppresses haze and glow effectively. |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
RoomTex: Texturing Compositional Indoor Scenes via Iterative Inpainting
## Method
In this section, we present our coarse-to-fine generation framework, RoomTex, for synthesizing high-fidelity and style-consistent texture for a compositional room. We utilize off-the-shelf 3D shape gener... | • Iterative Inpainting: Uses ControlNet with depth guidance to inpaint novel views of objects.
• Misalignment Removal: Detects and corrects misalignments between RGB and depth maps.
• Fine-grained texture control: Allows adding/removing/modifying textures interactively. |
Generate 4 bullet points for the "Research Motivation / Problem Background" section of an academic poster. | ## Paper Content
# Paper Title
TopoCellGen: Generating Histopathology Cell Topology with a Diffusion Model
## Abstract
Accurately modeling multi-class cell topology is crucial in digital pathology, as it provides critical insights into tissue structure and pathology. The synthetic generation of cell topology enable... | • Cell topology includes cell counts, intra-class distribution, and inter-class interactions.
• Diagnosis depends on who a cell's neighbors are.
• Current generative models lack accurate control of cell counts and ignore topology.
• We aim to generate cell topology accurately! |
Generate 3 bullet points for the "Background / Related Work" section of an academic poster. | ## Paper Content
# Paper Title
PETRA: PARALLEL END-TO-END TRAINING OF REVERSIBLE ARCHITECTURES
## Abstract
Reversible architectures have been shown to be capable of performing on par with their non-reversible architectures, being applied in deep learning for memory savings and generative modeling. In this work, we ... | • Invertible functions allow recomputing activations during backward pass without storing them.
• Formally: x_{j+1} = x̃_j + F_j(x_j) and x_j = x̃_{j+1} - F_j(x̃_{j+1}).
• Examples include RevNet, i-ResNet, and Reformer architectures. |
Generate 2 bullet points for the "Background / Related Work" section of an academic poster. | ## Paper Content
# Paper Title
ODE: Open-Set Evaluation of Hallucinations in Multimodal Large Language Models
## Abstract
Hallucination poses a persistent challenge for multimodal large language models (MLLMs). However, existing benchmarks for evaluating hallucinations are generally static, which may overlook the p... | • Images and scenes from dataset: rider and train, pedestrian and highway, train and motorcycle, bicycle and tunnel, train and gas station, pedestrian and trailer.
• Illustrates diversity of visual content used in evaluation. |
Generate 5 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
Adversarial Backdoor Attack by Naturalistic Data Poisoning on Trajectory Prediction in Autonomous Driving
## Abstract
In autonomous driving, behavior prediction is fundamental for safe motion planning, hence the security and robustness of prediction models against adversarial attacks ... | • Clip and perturb the gradient of the weights that are perceived abnormal to mitigate the effect of poisoned samples
• Defence's impact is minimal
• Smoothing without attack degrades performance
• With disguising attack remains effective
• Without disguising effectiveness drops drastically |
Generate 3 bullet points for the "Background / Related Work" section of an academic poster. | ## Paper Content
# Paper Title
Structure-from-Motion with a Non-Parametric Camera Model
## Abstract
In this paper, we present a new generic Structure-from-Motion pipeline, GenSfM, that uses a non-parametric camera projection model. The model is self-calibrated during the reconstruction process and can fit a wide va... | • Introduces Structure-from-Motion (SfM) and parametric camera models (Brown-Conrady, Kannala-Brandt).
• Highlights pitfalls of previous SfM methods when handling severe distortions like fisheye or catadioptric images.
• Notes that standard SfM relies on wisely chosen parametric camera models. |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
FLOW MATCHING WITH GENERAL DISCRETE PATHS: A KINETIC-OPTIMAL PERSPECTIVE
## Abstract
The design space of discrete-space diffusion or flow generative models are significantly less well-understood than their continuous-space counterparts, with many works focusing only on a simple masked... | • Solve the kinetic optimal problem for the probability path.
• The kinetic optimal velocity is constrained to the hypersphere.
• The solutions are the geodesics. |
Generate 3 bullet points for the "Method Overview / Framework" section of an academic poster. | ## Paper Content
# Paper Title
RAM-Avatar: Real-time Photo-Realistic Avatar from Monocular Videos with Full-body Control
## Abstract
This paper focuses on advancing the applicability of human avatar learning methods by proposing RAM-Avatar, which learns a Real-time, photo-realistic Avatar that supports full-body co... | • Proposes a real-time animatable monocular framework supporting photo-realistic full-body animation including body poses, hand gestures, and facial expressions.
• Introduces a dual attention module for stable realistic clothed body details by considering geometric misalignment.
• Includes a motion distribution alignme... |
Generate 2 bullet points for the "Background / Related Work" section of an academic poster. | ## Paper Content
# Paper Title
A Tale of Two Classes: Adapting Supervised Contrastive Learning to Binary Imbalanced Datasets
## Abstract
Supervised contrastive learning (SupCon) has proven to be a powerful alternative to the standard cross-entropy loss for classification of multi-class balanced datasets. However, i... | • Canonical metrics like Sample Alignment Distance (SAD) and Class Alignment Distance (CAD) measure within-class similarity.
• They can show high alignment even when classes are poorly separated (representation collapse). |
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster. | ## Paper Content
# Paper Title
REVISITING ZEROTH-ORDER OPTIMIZATION: MINIMUM-VARIANCE TWO-POINT ESTIMATORS AND DIRECTIONALLY ALIGNED PERTURBATIONS
## Abstract
In this paper, we explore the two-point zeroth-order gradient estimator and identify the distribution of random perturbations that minimizes the estimator's ... | • DAPs exhibit directional alignment property: effects vary by projection direction.
• Unlike uniform perturbations, DAPs adaptively introduce low-variance perturbations where the true gradient is large.
• This reduces noise in gradient estimation, especially when gradients are sparse or vary in magnitude. |
Generate 5 bullet points for the "Method Overview / Framework" section of an academic poster. | ## Paper Content
# Paper Title
Flowing Datasets with Wasserstein over Wasserstein Gradient Flows
## Abstract
Many applications in machine learning involve data represented as probability distributions. The emergence of such data requires radically novel techniques to design tractable gradient flows on probability d... | • Goal: move labeled dataset in a coherent way
• Labeled datasets modeled as P = 1/C Σ δ_{x^c} ∈ P₂(P₂(R^d))
• Endow P₂(P₂(R^d)) with OT distance WoW
• Minimize F: P₂(P₂(R^d)) → R using WoW gradient flows
• Application on image datasets |
Generate 4 bullet points for the "Method Overview / Framework" section of an academic poster. | ## Paper Content
# Paper Title
MobileMamba: Lightweight Multi-Receptive Visual Mamba Network
## Abstract
Previous research on lightweight models has primarily focused on CNNs and Transformer-based designs. CNNs, with their local receptive fields, struggle to capture long-range dependencies, while Transformers, desp... | • We propose MobileMamba, a lightweight framework balancing efficiency and performance.
• Introduces Multi-Receptive Field Feature Interaction (MRFFI) module with WTE-Mamba and MK-DeConv.
• Achieves 83.6% Top-1 accuracy, 21x faster than LocalVim on GPU.
• Extensive experiments show superior balance of speed and accurac... |
Generate 2 bullet points for the "Conclusion / Future Work" section of an academic poster. | ## Paper Content
# Paper Title
Towards an Information Theoretic Framework of Context-Based Offline Meta-Reinforcement Learning
## Abstract
As a marriage between offline RL and meta-RL, the advent of offline meta-reinforcement learning (OMRL) has shown great promise in enabling RL agents to multi-task and quickly ad... | • I(Z;M) operates as a unified learning objective and is robust to context shift.
• It achieves this by trading off the primary and lesser causalities of COMRL. |
Generate 3 bullet points for the "Research Motivation / Problem Background" section of an academic poster. | ## Paper Content
# Paper Title
FreeZe: Training-free zero-shot 6D pose estimation with geometric and vision foundation models
## Introduction
In our daily interactions, we easily manipulate objects around us, whether by grasping a mug or pouring water into a glass, thanks to our subconscious ability to locate them ... | • Eliminates need for training time and data for 6D pose estimation.
• Current methods require extensive training on specific objects or synthetic datasets.
• FreeZe offers a training-free, zero-shot approach using pre-trained foundation models. |
Generate 3 bullet points for the "Qualitative Results / Visualization" section of an academic poster. | ## Paper Content
# Paper Title
FSD-BEV: Foreground Self-Distillation for Multi-view 3D Object Detection
## Experiments
Dataset and Metrics We conducted our experiments on the nuScenes [1] dataset, which comprises 1000 sequences captured by sensors such as cameras and LiDAR. Each sequence is approximately 20 seconds... | • Visual comparison between backbone-only and our FSD-BEV method.
• Our method produces more accurate and complete 3D object detections.
• Green bounding boxes indicate correct detections; red indicate misses or false positives. |
Generate 2 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster. | ## Paper Content
# Paper Title
Cascade Prompt Learning for Vision-Language Model Adaptation
## Experiments
Datasets. For base-to-novel generalization and few-shot experiments, we use 11 datasets following [64, 65]. Specifically, the datasets include ImageNet [4] and Caltech101 [9] for generic objecting, FGVCAircraf... | • HM of prompt learning methods with or without our CasPL on base-to-novel tasks.
• CasPL performance comparison in a few-shot image recognition setting across 11 datasets. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.