CVPR 2024 Papers — Page 4
IEEE/CVF Conference on Computer Vision and Pattern Recognition · 2716 papers
Can't Make an Omelette Without Breaking Some Eggs: Plausible Action Anticipation Using Large Video-Language Models
Himangi Mittal (Carnegie Mellon University), Kwonjoon Lee (Carnegie Mellon University)
GenerationOptimizationTransformerLarge Language ModelVision Language ModelVideoText
🎯 What it does: PlausibleVL is designed, a large-scale video-language model for predicting feasible future action sequences in the real world, enhancing the model's temporal consistency and action diversity by introducing two novel loss functions.
CAPE: CAM as a Probabilistic Ensemble for Enhanced DNN Interpretation
Townim Faisal Chowdhury (Australian Institute for Machine Learning, University of Adelaide), Zhibin Liao (SA Pathology)
Explainability and InterpretabilityKnowledge DistillationConvolutional Neural NetworkTransformerImage
🎯 What it does: Proposes the CAPE method, which reinterprets CAM as a probabilistic ensemble for visualizing DNN attention areas and providing comparable explanations.
CapHuman: Capture Your Moments in Parallel Universes
Chao Liang (Zhejiang University), Yi Yang (Zhejiang University)
GenerationData SynthesisPose EstimationDiffusion modelImage
🎯 What it does: Proposes the CapHuman framework, which generates high-fidelity portraits under various poses, expressions, and lighting changes using a single reference face image.
CapsFusion: Rethinking Image-Text Data at Scale
Qiying Yu (Institute for AI Industry Research), Jingjing Liu (Institute for AI Industry Research)
GenerationData SynthesisTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodality
🎯 What it does: A framework named CAPSFUSION is proposed, which integrates raw web text and synthetic subtitles through large language models to generate high-quality multimodal data and train LMM;
Capturing Closely Interacted Two-Person Motions with Reaction Priors
Qi Fang (NetEase Games AI Lab), Kang Chen (NetEase Games AI Lab)
Object TrackingPose EstimationTransformerAuto EncoderVideoText
🎯 What it does: This work proposes a framework based on 'reaction priors' to capture close interaction actions between two people from monocular videos, achieving precise recovery of severely occluded interaction actions through a query-based Transformer estimator.
Carve3D: Improving Multi-view Reconstruction Consistency for Diffusion Models with RL Finetuning
Desai Xie (Adobe Research), Arie E. Kaufman (Stony Brook University)
GenerationData SynthesisReinforcement LearningDiffusion modelNeural Radiance FieldImagePoint Cloud
🎯 What it does: By performing reinforcement learning fine-tuning (RLFT) on a multi-view diffusion model and combining it with a new multi-view reconstruction consistency (MRC) metric, the perspective consistency of generated 3D models and the quality of NeRF reconstruction have been significantly improved.
CARZero: Cross-Attention Alignment for Radiology Zero-Shot Classification
Haoran Lai (University of Science and Technology of China), S. Kevin Zhou (University of Science and Technology of China)
ClassificationTransformerLarge Language ModelPrompt EngineeringContrastive LearningImageTextBiomedical Data
🎯 What it does: This work proposes the CARZero framework, which achieves zero-shot classification of chest X-ray images and radiology reports through cross-attention alignment methods and LLM-driven prompt alignment.
CAT-DM: Controllable Accelerated Virtual Try-on with Diffusion Model
Jianhao Zeng (Tianjin University), An-An Liu (Tianjin University)
GenerationData SynthesisComputational EfficiencyDiffusion modelGenerative Adversarial NetworkImage
🎯 What it does: A controllable accelerated virtual try-on model CAT-DM is proposed, which integrates ControlNet and the rough results of a pre-trained GAN, enhancing the fine reproduction of clothing textures while compressing the sampling steps of the diffusion model from the conventional hundreds to just 2 steps.
CAT-Seg: Cost Aggregation for Open-Vocabulary Semantic Segmentation
Seokju Cho (Korea University), Seungryong Kim (Korea University)
SegmentationTransformerVision Language ModelImageMultimodality
🎯 What it does: This paper proposes a cost aggregation-based open-window vocabulary semantic segmentation framework (CAT-Seg), which effectively transfers the global semantic knowledge of CLIP to pixel-level tasks by aggregating the cosine similarity (cost volume) of CLIP image and text embeddings and performing spatial and category aggregation. It also efficiently fine-tunes the CLIP encoder, making the model applicable to both seen and unseen classes.
CAT: Exploiting Inter-Class Dynamics for Domain Adaptive Object Detection
Mikhail Kennerley (National University of Singapore), Robby T. Tan (National University of Singapore)
Object DetectionDomain AdaptationKnowledge DistillationImage
🎯 What it does: A Class-Aware Teacher (CAT) framework is proposed to address the class imbalance problem in object detection within unlabeled target domains.
Category-Level Multi-Part Multi-Joint 3D Shape Assembly
Yichen Li (Stanford University), Lin Shao (National University of Singapore)
OptimizationGraph Neural NetworkPoint Cloud
🎯 What it does: This paper proposes a hierarchical graph network for multi-part and multi-joint 3D shape assembly at the category level, capable of simultaneously optimizing shape structure and joint matching.
Causal Mode Multiplexer: A Novel Framework for Unbiased Multispectral Pedestrian Detection
Taeheon Kim (KAIST), Yong Man Ro (KAIST)
Object DetectionImage
🎯 What it does: This paper proposes a Causal Mode Multiplexer (CMM) framework that utilizes causal intervention and counterfactual inference to address the modality bias problem in multispectral pedestrian detection.
Causal-CoG: A Causal-Effect Look at Context Generation for Boosting Multi-modal Language Models
Shitian Zhao (East China Normal University), Yan Wang (East China Normal University)
RecognitionGenerationTransformerPrompt EngineeringVision Language ModelMultimodality
🎯 What it does: A training-free, context-based generative reasoning strategy called Causal-CoG is proposed for multimodal language models to answer visual question answering problems.
CausalPC: Improving the Robustness of Point Cloud Classification by Causal Effect Identification
Yuanmin Huang (Fudan University), Min Yang (Fudan University)
ClassificationAdversarial AttackPoint Cloud
🎯 What it does: A point cloud classification defense framework based on causal reasoning, CausalPC, is proposed, which utilizes structural information and reconstructed point clouds to jointly infer causal effects, enhancing adversarial robustness.
CCEdit: Creative and Controllable Video Editing via Diffusion Models
Ruoyu Feng (University of Science and Technology of China), Baining Guo (Microsoft Research Asia)
GenerationData SynthesisDiffusion modelVideoBenchmark
🎯 What it does: A multifunctional video editing framework CCEdit based on diffusion models is proposed, capable of performing editing tasks such as style transfer and foreground/background replacement through structural control (ControlNet) and appearance control (keyframe editing).
CDFormer: When Degradation Prediction Embraces Diffusion Model for Blind Image Super-Resolution
Qingguo Liu (Nanjing University of Aeronautics and Astronautics), Jie Qin (Nanjing University of Aeronautics and Astronautics)
RestorationSuper ResolutionTransformerDiffusion modelImage
🎯 What it does: A content-aware degradation-driven transformation network (CDFormer) for blind image super-resolution is proposed, which utilizes a diffusion model to estimate the content degradation prior (CDP) of low-resolution images and achieves efficient feature fusion through attention and interflow mechanisms, ultimately generating high-quality super-resolved images.
CDMAD: Class-Distribution-Mismatch-Aware Debiasing for Class-Imbalanced Semi-Supervised Learning
Hyuck Lee (Korea Advanced Institute of Science and Technology), Heeyoung Kim (Korea Advanced Institute of Science and Technology)
ClassificationRepresentation LearningSupervised Fine-TuningImage
🎯 What it does: The CDMAD algorithm is proposed to address the issue of mismatched class distribution between labeled and unlabeled data in class-imbalanced semi-supervised learning. It utilizes unpatterned images to compute classifier bias and corrects pseudo-labels and test predictions, thereby reducing classifier bias and improving representation quality.
CFAT: Unleashing Triangular Windows for Image Super-resolution
Abhisek Ray (Indian Institute of Technology Patna), Maheshkumar H. Kolekar (Indian Institute of Technology Patna)
RestorationSuper ResolutionTransformerImage
🎯 What it does: This paper proposes an image super-resolution model named Composite Fusion Attention Transformer (CFAT), which utilizes a self-attention mechanism that combines triangular and rectangular windows to enhance reconstruction quality.
CFPL-FAS: Class Free Prompt Learning for Generalizable Face Anti-spoofing
Ajian Liu (Chinese Academy of Sciences Institute of Automation), Zhen Lei (Chinese Academy of Sciences Institute of Automation)
RecognitionDomain AdaptationTransformerPrompt EngineeringContrastive LearningImage
🎯 What it does: A Class Free Prompt Learning (CFPL) framework based on CLIP is proposed, utilizing two lightweight Q-Formers to generate content and style prompts, and achieving cross-domain generalization for face anti-spoofing through text supervision and style diversification.
CG-HOI: Contact-Guided 3D Human-Object Interaction Generation
Christian Diller (Technical University of Munich), Angela Dai (Technical University of Munich)
GenerationPose EstimationDiffusion modelTextPoint Cloud
🎯 What it does: This paper proposes a method for generating realistic 3D human-object interaction sequences from text descriptions and static object geometries, jointly predicting human poses, object movements, and contact distances;
CGI-DM: Digital Copyright Authentication for Diffusion Models via Contrasting Gradient Inversion
Xiaoyu Wu (Shanghai Jiao Tong University), Haibing Guan (Shanghai Jiao Tong University)
GenerationData SynthesisOptimizationDiffusion modelContrastive LearningImage
🎯 What it does: A digital copyright identification method based on contrast gradient inversion, CGI-DM, is proposed, which utilizes partial image information recovery to verify whether the diffusion model has memorized the training samples.
ChAda-ViT : Channel Adaptive Attention for Joint Representation Learning of Heterogeneous Microscopy Images
Nicolas Bourriez (Ecole Normale Superieure PSL), Auguste Genovesio (Ecole Normale Superieure PSL)
Representation LearningTransformerContrastive LearningImageBiomedical Data
🎯 What it does: A Transformer architecture called ChAda-ViT is proposed, capable of handling microscope images with arbitrary channel numbers and types, achieving a unified embedding space.
CHAIN: Enhancing Generalization in Data-Efficient GANs via lipsCHitz continuity constrAIned Normalization
Yao Ni (Australian National University), Piotr Koniusz (Australian National University)
GenerationData SynthesisGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes CHAIN (lipsCHitz continuity constrAIned Normalization), an improved normalization method specifically designed for GAN discriminators to enhance generalization and training stability in data-scarce scenarios.
Characteristics Matching Based Hash Codes Generation for Efficient Fine-grained Image Retrieval
Zhen-Duo Chen (Shandong University), Xin-Shun Xu (Shandong University)
RetrievalKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: A feature matching-based hash code generation method called CMBH is proposed to address the contradiction between feature learning and hash code capacity, as well as the conflict between efficiency and effectiveness in fine-grained image retrieval.
Chat-UniVi: Unified Visual Representation Empowers Large Language Models with Image and Video Understanding
Peng Jin (Peking University), Li Yuan (Peking University)
GenerationRepresentation LearningTransformerLarge Language ModelVision Language ModelImageVideoMultimodality
🎯 What it does: This paper proposes and implements Chat-UniVi, a unified visual-language model capable of understanding images and videos within the same framework and engaging in natural dialogue with large language models.
ChatPose: Chatting about 3D Human Pose
Yao Feng (Max Planck Institute for Intelligent Systems), Michael J. Black (Max Planck Institute for Intelligent Systems)
GenerationPose EstimationTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelTextMultimodality
🎯 What it does: Train a multimodal LLM ChatPose to understand and generate SMPL 3D human poses, supporting chat-based interactions based on text or images.
ChatScene: Knowledge-Enabled Safety-Critical Scenario Generation for Autonomous Vehicles
Jiawei Zhang (University of Illinois Urbana-Champaign), Bo Li (University of Chicago)
Autonomous DrivingSafty and PrivacyTransformerLarge Language ModelReinforcement LearningTextRetrieval-Augmented Generation
🎯 What it does: This paper proposes the ChatScene agent based on large language models, which can automatically generate natural language descriptions of safety-critical scenarios and convert them into Scenic code that can be executed in the CARLA simulation environment, thereby producing diverse and complex safety testing scenarios.
Check Locate Rectify: A Training-Free Layout Calibration System for Text-to-Image Generation
Biao Gong (Alibaba Group), Yu Liu (Alibaba Group)
GenerationData SynthesisTransformerDiffusion modelImageText
🎯 What it does: A training-independent layout correction system called SimM is proposed, which automatically detects and corrects layout inconsistencies during text-to-image generation.
Choose What You Need: Disentangled Representation Learning for Scene Text Recognition Removal and Editing
Boqiang Zhang (University of Science and Technology of China), Yuxin Wang (University of Science and Technology of China)
RecognitionRepresentation LearningTransformerImage
🎯 What it does: The DARLING framework is proposed to achieve decoupled representation learning for scene text multi-tasks (recognition, editing, removal).
Cinematic Behavior Transfer via NeRF-based Differentiable Filming
Xuekun Jiang (Shanghai AI Lab), Bo Dai (Shanghai AI Lab)
Pose EstimationOptimizationNeural Radiance FieldSimultaneous Localization and MappingVideo
🎯 What it does: This paper proposes a method for estimating inverse shooting behavior, utilizing NeRF as a differentiable renderer to optimize camera trajectories and refine SMPL poses, followed by behavior transfer of 2D/3D film shots through a 3D engine.
Circuit Design and Efficient Simulation of Quantum Inner Product and Empirical Studies of Its Effect on Near-Term Hybrid Quantum-Classic Machine Learning
Hao Xiong (Shanghai Jiao Tong University), Junchi Yan (Shanghai Jiao Tong University)
ClassificationOptimizationComputational EfficiencyGraph Neural NetworkReinforcement LearningImageGraph
🎯 What it does: This paper designs quantum inner product (QIP) circuits suitable for 1-to-1, 1-to-N, and M-to-N scenarios, and proposes an efficient scheme for directly simulating output states, enabling rapid computation of quantum inner products on classical machines while maintaining differentiability.
CityDreamer: Compositional Generative Model of Unbounded 3D Cities
Haozhe Xie (Nanyang Technological University), Ziwei Liu (Nanyang Technological University)
GenerationData SynthesisGenerative Adversarial NetworkImage
🎯 What it does: This paper presents CityDreamer, a generative model capable of generating unbounded 3D cities.
Class Incremental Learning with Multi-Teacher Distillation
Haitao Wen (University of Electronic Science and Technology of China), Hongliang Li (University of Electronic Science and Technology of China)
ClassificationKnowledge DistillationImage
🎯 What it does: This paper proposes a multi-teacher distillation framework (MTD) to alleviate catastrophic forgetting in class-incremental learning by generating diverse teachers from a single model.
Class Tokens Infusion for Weakly Supervised Semantic Segmentation
Sung-Hoon Yoon (Korea Advanced Institute of Science and Technology), Kuk-Jin Yoon (Korea Advanced Institute of Science and Technology)
SegmentationTransformerImage
🎯 What it does: A weakly supervised semantic segmentation framework based on Vision Transformer is proposed, which enhances the distinguishability of class tokens and the quality of CAM by injecting class tokens (I-CTI and C-CTI) and background tokens into the Transformer.
Classes Are Not Equal: An Empirical Study on Image Recognition Fairness
Jiequan Cui (Nanyang Technological University), Hanwang Zhang (Nanyang Technological University)
ClassificationRecognitionRepresentation LearningConvolutional Neural NetworkTransformerContrastive LearningImage
🎯 What it does: A systematic empirical study on the fairness of image classification models (i.e., extreme accuracy differences among different categories) is conducted, verifying the universality of this issue across various datasets, network architectures, and model capacities.
CLIB-FIQA: Face Image Quality Assessment with Confidence Calibration
Fu-Zhao Ou (City University of Hong Kong), Sam Kwong (Lingnan University)
ClassificationRecognitionVision Language ModelContrastive LearningImage
🎯 What it does: This paper utilizes the CLIP visual-language pre-training model to jointly learn multiple quality factors (blurriness, pose, expression, occlusion, lighting) and quality distribution, proposing a confidence-calibrated facial image quality assessment method (CLIB-FIQA) that achieves accurate prediction of facial image quality and calibrates the confidence of quality anchors.
CLiC: Concept Learning in Context
Mehdi Safaee (Simon Fraser University), Ali Mahdavi-Amiri (Simon Fraser University)
GenerationData SynthesisDiffusion modelImage
🎯 What it does: Learning local visual concepts (such as chair decoration) from a single image and transferring them to a target image or generating new images while maintaining the geometric and structural features of the concept in its original context.
CLIP as RNN: Segment Countless Visual Concepts without Training Endeavor
Shuyang Sun (University of Oxford), Siyang Li (Google Research)
Object DetectionSegmentationRecurrent Neural NetworkVision Language ModelContrastive LearningImage
🎯 What it does: This paper studies a zero-shot open vocabulary image segmentation method that utilizes a pre-trained CLIP model to achieve semantic and reference segmentation without any mask annotations or additional training.
CLIP-BEVFormer: Enhancing Multi-View Image-Based BEV Detector with Ground Truth Flow
Chenbin Pan (Syracuse University), Liu Ren (Bosch Research North America)
Object DetectionAutonomous DrivingTransformerContrastive LearningImage
🎯 What it does: This paper proposes a framework called CLIP-BEVFormer, which introduces ground truth flow (GT-flow) guidance during the training process of multi-view image BEV detectors, significantly enhancing BEV representation and detection performance.
CLIP-Driven Open-Vocabulary 3D Scene Graph Generation via Cross-Modality Contrastive Learning
Lianggangxu Chen (East China Normal University), Gaoqi He (East China Normal University)
Object DetectionGenerationContrastive LearningMultimodalityPoint Cloud
🎯 What it does: An open-source semantic 3D scene graph generation framework for cross-modal contrastive learning is proposed, capable of achieving open vocabulary/zero-shot prediction without labeled data.
CLIP-KD: An Empirical Study of CLIP Model Distillation
Chuanguang Yang (Institute of Computing Technology), Yongjun Xu (Institute of Computing Technology)
ClassificationRetrievalKnowledge DistillationTransformerContrastive LearningImageTextMultimodality
🎯 What it does: This paper proposes a knowledge distillation framework for the CLIP model called CLIP-KD, which uses a large teacher CLIP to train a small student CLIP, significantly improving the zero-shot classification and cross-modal retrieval performance of the small model.
CLIPtone: Unsupervised Learning for Text-based Image Tone Adjustment
Hyeongmin Lee (POSTECH), Sunghyun Cho (POSTECH)
Image TranslationImage HarmonizationContrastive LearningImageText
🎯 What it does: This paper proposes CLIPtone, an unsupervised learning framework for text-driven image tone adjustment that can adaptively adjust the tone of images based on natural language descriptions.
CLOAF: CoLlisiOn-Aware Human Flow
Andrey Davydov (École Polytechnique Fédérale de Lausanne), Pascal Fua (École Polytechnique Fédérale de Lausanne)
Pose EstimationVideoOrdinary Differential Equation
🎯 What it does: This paper proposes a motion field integration method based on differentiable ODEs (CLOAF), which can completely eliminate the self-intersection problem in 3D human pose estimation while maintaining the prior of human shape, and can be used for model fine-tuning.
Clockwork Diffusion: Efficient Generation With Model-Step Distillation
Amirhossein Habibian (Qualcomm AI Research), Jens Petersen (Qualcomm AI Research)
GenerationData SynthesisComputational EfficiencyKnowledge DistillationConvolutional Neural NetworkDiffusion modelImageText
🎯 What it does: This paper proposes Clockwork Diffusion, which achieves efficient sampling for text-to-image generation and image editing by combining model distillation and step distillation on low-resolution UNet representations.
Close Imitation of Expert Retouching for Black-and-White Photography
Seunghyun Shin (Gwangju Institute of Science and Technology), Hae-Gon Jeon (Inha University)
Image TranslationRestorationConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: A specialized reproduction of black and white photography is proposed based on deep metric learning with a proxy network and multi-level bilateral grid.
Closely Interactive Human Reconstruction with Proxemics and Physics-Guided Adaption
Buzhen Huang (National University of Singapore), Gim Hee Lee (National University of Singapore)
GenerationPose EstimationTransformerDiffusion modelVideo
🎯 What it does: This paper proposes a dual-branch diffusion model that combines social proxemics priors and physical constraints to reconstruct the body posture and shape of two people in close interaction using visual information from monocular videos.
Cloud-Device Collaborative Learning for Multimodal Large Language Models
Guanqun Wang (Peking University), Shanghang Zhang (Peking University)
CompressionDomain AdaptationKnowledge DistillationTransformerLarge Language ModelTextMultimodality
🎯 What it does: A cloud-device collaborative continuous adaptation framework (CD-CCA) is proposed to achieve continuous adaptation of compressed multimodal large language models on the device side in dynamic environments.
CLOVA: A Closed-LOop Visual Assistant with Tool Usage and Update
Zhi Gao (Peking University), Qing Li
OptimizationTransformerLarge Language ModelPrompt EngineeringImageTextMultimodality
🎯 What it does: A closed-loop visual assistant called CLOVA is proposed, which continuously updates LLMs and visual tools during the inference, reflection, and learning stages to enhance multimodal task performance.
Clustering for Protein Representation Learning
Ruijie Quan (Zhejiang University), Yi Yang (Zhejiang University)
Representation LearningDrug DiscoveryGraph Neural NetworkGraphBiomedical Data
🎯 What it does: An end-to-end learnable neural clustering framework is proposed, which automatically discovers key amino acids in proteins through iterative spherical clustering, attention clustering representation, and GCN selection of key clusters, thereby achieving protein representation learning.
Clustering Propagation for Universal Medical Image Segmentation
Yuhang Ding (University of Technology Sydney), Yi Yang (Zhejiang University)
SegmentationTransformerImageBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: A unified medical image segmentation framework S2VNet is proposed, capable of achieving both automatic and interactive voxel segmentation in a single model and single training session.
CMA: A Chromaticity Map Adapter for Robust Detection of Screen-Recapture Document Images
Changsheng Chen (Shenzhen University), Jiwu Huang (Shenzhen University)
RecognitionAnomaly DetectionTransformerSupervised Fine-TuningImage
🎯 What it does: Designed and implemented a chromatic map-based adapter (CMA) for detecting screen recaptured document images.
CN-RMA: Combined Network with Ray Marching Aggregation for 3D Indoor Object Detection from Multi-view Images
Guanlin Shen, Bin Wang
RecognitionObject DetectionConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a new computer vision method aimed at improving the accuracy of image recognition.
CNC-Net: Self-Supervised Learning for CNC Machining Operations
Mohsen Yavartanoo (Seoul National University), Kyoung Mu Lee (Seoul National University)
Convolutional Neural NetworkRecurrent Neural NetworkPoint CloudMesh
🎯 What it does: A self-supervised CNC-Net is proposed, which predicts the sequence of milling, drilling, and rotation operations required by CNC machines from 3D models, gradually processing raw materials to obtain the target object.
Co-Speech Gesture Video Generation via Motion-Decoupled Diffusion Model
Xu He (Shenzhen International Graduate School Tsinghua University), Xiaofei Wu (Huawei)
GenerationData SynthesisTransformerDiffusion modelOptical FlowVideoMultimodalityAudio
🎯 What it does: This work proposes an end-to-end motion decoupling framework for generating synchronized co-speech gesture videos with audio.
Coarse-to-Fine Latent Diffusion for Pose-Guided Person Image Synthesis
Yanzuo Lu (Sun Yat-sen University), Jianhuang Lai (Sun Yat-sen University)
GenerationData SynthesisPose EstimationDiffusion modelImage
🎯 What it does: A coarse-to-fine staged latent diffusion model CFLD is proposed for pose-guided portrait image synthesis.
COCONut: Modernizing COCO Segmentation
Xueqing Deng (ByteDance), Liang-Chieh Chen (ByteDance)
Object DetectionSegmentationTransformerImage
🎯 What it does: Created the COCONut dataset, which contains 383K images and 5.18M high-quality, manually verified segmentation masks, uniformly covering three tasks: semantic, instance, and panoptic segmentation.
CoDe: An Explicit Content Decoupling Framework for Image Restoration
Enxuan Gu (Dalian University of Technology), Yong Guo (South China University of Technology)
RestorationSuper ResolutionImage
🎯 What it does: This paper proposes the CoDe framework, which achieves end-to-end image restoration through frequency domain content decoupling, sub-network specialization, and consistency loss.
Codebook Transfer with Part-of-Speech for Vector-Quantized Image Modeling
Baoquan Zhang (Harbin Institute of Technology), Yao He (ShenZhen SiFar Co., Ltd.)
GenerationData SynthesisGraph Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: A codebook transfer framework VQCT based on the vocabulary of pre-trained language models and part-of-speech knowledge is proposed to transfer semantically rich codebooks to VQIM and generate them through graph convolutional networks, thereby alleviating the codebook collapse problem.
CodedEvents: Optimal Point-Spread-Function Engineering for 3D-Tracking with Event Cameras
Sachin Shah, Christopher A. Metzler
Object TrackingOptimizationConvolutional Neural NetworkSimultaneous Localization and MappingPoint Cloud
🎯 What it does: This paper derives the Cramér-Rao bound for event cameras, designs optimizable phase and amplitude masks, and utilizes implicit neural representation to achieve efficient PSF engineering, enhancing 3D point tracking performance.
CoDeF: Content Deformation Fields for Temporally Consistent Video Processing
Hao Ouyang (Hong Kong University of Science and Technology), Yujun Shen (Ant Group)
RestorationObject TrackingSuper ResolutionOptical FlowVideo
🎯 What it does: This paper proposes a Content Deformation Field (CoDeF) video representation, which splits the video into a two-dimensional content field and a three-dimensional temporal deformation field, achieving high-quality video reconstruction and consistent video processing across frames.
CoDi-2: In-Context Interleaved and Interactive Any-to-Any Generation
Zineng Tang (University of California Berkeley), Mohit Bansal (University of North Carolina Chapel Hill)
GenerationData SynthesisTransformerLarge Language ModelDiffusion modelImageVideoTextMultimodalityAudio
🎯 What it does: CoDi-2 is proposed, a multimodal large language model capable of handling any modality input and generating any modality output, supporting cross-modal instructions, contextual learning, and multi-turn dialogue.
CoDi: Conditional Diffusion Distillation for Higher-Fidelity and Faster Image Generation
Kangfu Mei (Google Research), Peyman Milanfar (Google Research)
RestorationGenerationSuper ResolutionKnowledge DistillationDiffusion modelImage
🎯 What it does: A method has been developed to directly distill a conditional diffusion model from an unconditional diffusion model, enabling the generation of high-quality images within just 1-4 sampling steps.
CoG-DQA: Chain-of-Guiding Learning with Large Language Models for Diagram Question Answering
Shaowei Wang (Xi'an Jiaotong University), Jun Liu (Xi'an Jiaotong University)
Graph Neural NetworkTransformerLarge Language ModelPrompt EngineeringImageTextMultimodality
🎯 What it does: This paper proposes a diagram-based question answering framework called CoG-DQA, which utilizes the collaboration between LLM and diagram analysis tools to generate multi-level visual features and semantic relationships, significantly enhancing the performance of diagram-based question answering.
CogAgent: A Visual Language Model for GUI Agents
Wenyi Hong (Tsinghua University), Jie Tang (Tsinghua University)
TransformerVision Language ModelImageTextMultimodality
🎯 What it does: A visual language model called CogAgent, focused on GUI understanding and navigation with 18 billion parameters, has been developed, and high-resolution input support has been implemented.
CoGS: Controllable Gaussian Splatting
Heng Yu (Carnegie Mellon University), László A. Jeni (Carnegie Mellon University)
GenerationData SynthesisOptimizationGaussian SplattingVideo
🎯 What it does: A controllable 3D Gaussian split model is established using monocular video, enabling dynamic scene reconstruction and real-time manipulation;
Coherence As Texture - Passive Textureless 3D Reconstruction by Self-interference
Wei-Yu Chen (Carnegie Mellon University), Matthew O'Toole (Carnegie Mellon University)
Depth EstimationImage
🎯 What it does: A completely passive texture-independent 3D reconstruction method is proposed, which generates 'texture' in a self-interference device by utilizing the spatial coherence of natural light sources, thereby achieving depth estimation on surfaces without texture.
Coherent Temporal Synthesis for Incremental Action Segmentation
Guodong Ding (National University of Singapore), Angela Yao (National University of Singapore)
SegmentationGenerationData SynthesisAuto EncoderVideo
🎯 What it does: This paper proposes an incremental temporal action segmentation method based on generative models to address the catastrophic forgetting problem in the video domain.
Collaborating Foundation Models for Domain Generalized Semantic Segmentation
Yasser Benigmim (Telecom Paris), Stéphane Lathuilière (Telecom Paris)
SegmentationDomain AdaptationLarge Language ModelDiffusion modelImage
🎯 What it does: The CLOUDS framework is proposed to collaboratively achieve domain generalization semantic segmentation using multiple foundational models: first, robust features are extracted using a CLIP backbone, then diverse text prompts are generated through an LLM to drive a diffusion model to synthesize training images with varied content, followed by self-training and SAM to refine pseudo-labels, ultimately resulting in a segmentation model that performs excellently on unseen domains.
Collaborative Learning of Anomalies with Privacy (CLAP) for Unsupervised Video Anomaly Detection: A New Baseline
Anas Al-lahham (Mohamed bin Zayed University of Artificial Intelligence), Karthik Nandakumar (Mohamed bin Zayed University of Artificial Intelligence)
Anomaly DetectionFederated LearningSafty and PrivacyConvolutional Neural NetworkTransformerVideo
🎯 What it does: A federated learning framework called CLAP is proposed, which can train video anomaly detection models in a multi-participant environment in a completely unsupervised manner, while not sharing the original video data.
Collaborative Semantic Occupancy Prediction with Hybrid Feature Fusion in Connected Automated Vehicles
Rui Song (Fraunhofer IVI), Alois Knoll (Technical University of Munich)
Object DetectionSegmentationAutonomous DrivingConvolutional Neural NetworkPoint Cloud
🎯 What it does: A camera-based collaborative semantic occupancy prediction framework, CoHFF, is proposed, achieving 3D semantic occupancy prediction through feature sharing among multiple vehicles.
COLMAP-Free 3D Gaussian Splatting
Yang Fu (University of California San Diego), Xiaolong Wang (University of California Berkeley)
Pose EstimationOptimizationNeural Radiance FieldGaussian SplattingSimultaneous Localization and MappingVideoPoint Cloud
🎯 What it does: A framework based on 3D Gaussian Splatting (CF‑3DGS) is proposed, which achieves end-to-end joint optimization of camera pose estimation and scene reconstruction without relying on SfM for camera pose estimation, thus enabling new view rendering.
Color Shift Estimation-and-Correction for Image Enhancement
Yiyu Li (City University of Hong Kong), Rynson W.H. Lau (City University of Hong Kong)
RestorationConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: A method for color distortion estimation and correction for images containing both overexposed and underexposed regions is proposed, utilizing pseudo-normal exposure features to guide color shift correction and generate enhanced images.
ColorPCR: Color Point Cloud Registration with Multi-Stage Geometric-Color Fusion
Juncheng Mu (Tsinghua University), Yue Gao (Tsinghua University)
RecognitionRetrievalOptimizationTransformerPoint Cloud
🎯 What it does: A point cloud registration framework called ColorPCR is proposed, which achieves high-precision registration through multi-stage geometric-color fusion.
Combining Frame and GOP Embeddings for Neural Video Representation
Jens Eirik Saethre (Disney Research Studios), Christopher Schroers (Disney Research Studios)
CompressionRepresentation LearningOptical FlowVideo
🎯 What it does: A hybrid implicit neural representation model T-NeRV is proposed, which combines frame-level embeddings and GOP-level features to achieve adaptive video compression.
CommonCanvas: Open Diffusion Models Trained on Creative-Commons Images
Aaron Gokaslan (Cornell University), Volodymyr Kuleshov (Cornell University)
GenerationOptimizationComputational EfficiencyTransformerDiffusion modelImageText
🎯 What it does: A series of text-to-image diffusion models called CommonCanvas was trained using only Creative Commons images and synthetic captions, achieving generation quality comparable to Stable Diffusion 2 (SD2).
Commonsense Prototype for Outdoor Unsupervised 3D Object Detection
Hai Wu (Xiamen University), Cheng Wang (Xiamen University)
Object DetectionAutonomous DrivingPoint Cloud
🎯 What it does: This paper proposes an unsupervised 3D object detection framework called CPD based on consensus prototypes, which generates pseudo-labels through multi-frame clustering and refines the labels using consensus prototypes, ultimately achieving high-precision detection in sparse scanning environments.
Communication-Efficient Collaborative Perception via Information Filling with Codebook
Yue Hu (Shanghai Jiao Tong University), Siheng Chen (Shanghai Jiao Tong University)
Object DetectionOptimizationPoint Cloud
🎯 What it does: This paper proposes CodeFilling, a communication-efficient collaborative perception system that utilizes codebook compression and information filling techniques for 3D object detection.
Communication-Efficient Federated Learning with Accelerated Client Gradient
Geeho Kim (Seoul National University), Bohyung Han (Seoul National University)
OptimizationFederated LearningConvolutional Neural NetworkImage
🎯 What it does: Proposes FedACG, which utilizes the server to broadcast a global model with lookahead gradients as local initialization, and adds a regularization term to accelerate and stabilize federated learning.
Compact 3D Gaussian Representation for Radiance Field
Joo Chan Lee (Sungkyunkwan University), Eunbyung Park (Sungkyunkwan University)
GenerationCompressionNeural Radiance FieldGaussian SplattingPoint Cloud
🎯 What it does: A compressed version of the 3D Gaussian representation method is proposed, which reduces the number of Gaussian points and compresses attributes while maintaining high-quality scene reconstruction.
Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods
Mingqi Jiang (Oregon State University), Li Fuxin (Oregon State University)
ClassificationRecognitionExplainability and InterpretabilityKnowledge DistillationConvolutional Neural NetworkTransformerImage
🎯 What it does: This paper proposes two global statistical methods based on explanation algorithms—sub-explanation counting and cross-testing—for systematically evaluating the compositionality and disjunctivism of different visual recognition backbone networks in the decision-making process, as well as the similarity of feature usage.
Complementing Event Streams and RGB Frames for Hand Mesh Reconstruction
Jianping Jiang (Peking University), Boxin Shi (Peking University)
Pose EstimationTransformerImageMultimodalityMesh
🎯 What it does: This paper proposes EvRGBHand, a method for 3D hand mesh reconstruction that utilizes the fusion of event cameras and RGB cameras.
Composed Video Retrieval via Enriched Context and Discriminative Embeddings
Omkar Thawakar (Australian National University), Fahad Shahbaz Khan (Australian National University)
RetrievalTransformerLarge Language ModelContrastive LearningVideoTextMultimodality
🎯 What it does: A framework has been constructed that utilizes detailed language descriptions of query videos and learns multimodal discriminative embeddings to enhance the performance of Combined Video Retrieval (CoVR) and Combined Image Retrieval (CoIR).
Composing Object Relations and Attributes for Image-Text Matching
Khoi Pham (University of Maryland), Abhinav Shrivastava (University of Maryland)
RetrievalGraph Neural NetworkContrastive LearningImageText
🎯 What it does: A dual-encoder image-text matching model called CORA based on scene graphs is proposed.
Compositional Chain-of-Thought Prompting for Large Multimodal Models
Chancharik Mitra (University of California), Roei Herzig (University of California)
RecognitionGenerationTransformerLarge Language ModelPrompt EngineeringVision Language ModelMultimodalityBenchmarkChain-of-Thought
🎯 What it does: Using zero-shot chain-of-thought prompts to enhance the combinatorial reasoning ability of large multimodal models through generated scene graphs.
Compositional Video Understanding with Spatiotemporal Structure-based Transformers
Hoyeoung Yun (Hanyang University), Eun-Sol Kim (Hanyang University)
RecognitionObject DetectionTransformerVideo
🎯 What it does: This study proposes a spatiotemporal structure Transformer (ST-GT) based on objects, which can construct symbolic graphs from long videos and learn higher-order semantic relationships between object-level nodes and edges, achieving the decomposition and reconstruction of overall video semantics.
Compressed 3D Gaussian Splatting for Accelerated Novel View Synthesis
Simon Niedermayr (Technical University of Munich), Rüdiger Westermann (Technical University of Munich)
Data SynthesisCompressionComputational EfficiencyGaussian SplattingPoint Cloud
🎯 What it does: A compression method for 3D Gaussian splat scene representation is proposed, achieving low memory and real-time rendering through hardware rasterization.
Concept Weaver: Enabling Multi-Concept Fusion in Text-to-Image Models
Gihyun Kwon (KAIST), Fabian Caba Heilbron (Adobe)
GenerationData SynthesisDiffusion modelImage
🎯 What it does: A method called Concept Weaver is proposed for integrating multiple personalized concepts into text-to-image generation during the inference phase. It first generates a template image that aligns with the prompt semantics, and then injects the appearance of multiple concepts onto the template using strategies such as mask guidance, feature injection, and concept-independent suppression.
ConCon-Chi: Concept-Context Chimera Benchmark for Personalized Vision-Language Tasks
Andrea Rosasco (University of Genoa), Lorenzo Natale (Istituto Italiano di Tecnologia)
GenerationRetrievalTransformerVision Language ModelContrastive LearningImageTextMultimodalityBenchmark
🎯 What it does: Designed and released the ConCon-Chi dataset, constructing a concept-context matrix containing 20 novel 'hybrid' concepts and 101 diverse contexts, aimed at evaluating the learning of new meanings and combinatorial capabilities of personalized text-to-image retrieval (TIR) and generation (TIG) models.
Condition-Aware Neural Network for Controlled Image Generation
Han Cai (Massachusetts Institute of Technology), Song Han (NVIDIA)
GenerationData SynthesisConvolutional Neural NetworkTransformerDiffusion modelImageText
🎯 What it does: A Condition-Aware Neural Network (CAN) is proposed, which achieves controllable image generation for diffusion generative models by dynamically generating and integrating condition-aware weights.
CONFORM: Contrast is All You Need for High-Fidelity Text-to-Image Diffusion Models
Tuna Han Salih Meral (Virginia Tech), Pinar Yanardag (Virginia Tech)
GenerationData SynthesisOptimizationTransformerDiffusion modelContrastive LearningImageText
🎯 What it does: A training-independent, contrastive learning-based optimization method is proposed to enhance the fidelity of text-to-image diffusion models to multi-concept prompts.
Confronting Ambiguity in 6D Object Pose Estimation via Score-Based Diffusion on SE(3)
Tsu-Ching Hsiao (National Tsing Hua University), Chun-Yi Lee (National Tsing Hua University)
Pose EstimationDiffusion modelScore-based ModelImage
🎯 What it does: This paper proposes modeling the SE(3) Lie group using a score-based diffusion model on a single RGB image, thereby estimating both rotation and translation simultaneously, addressing the pose ambiguity caused by symmetric objects and occlusion.
ConsistDreamer: 3D-Consistent 2D Diffusion for High-Fidelity Scene Editing
Jun-Kun Chen (University of Illinois Urbana-Champaign), Yu-Xiong Wang (University of Illinois Urbana-Champaign)
GenerationData SynthesisDiffusion modelNeural Radiance FieldImagePoint Cloud
🎯 What it does: Proposes the ConsistDreamer framework, which utilizes a 2D diffusion model to achieve consistent high-fidelity editing of 3D scenes.
Consistency and Uncertainty: Identifying Unreliable Responses From Black-Box Vision-Language Models for Selective Visual Question Answering
Zaid Khan (Northeastern University), Yun Fu (Northeastern University)
RecognitionGenerationTransformerVision Language ModelMultimodality
🎯 What it does: The study identifies unreliable answers in black-box visual language models through neighborhood consistency and achieves selective prediction in visual question answering.
Consistent Prompting for Rehearsal-Free Continual Learning
Zhanxin Gao (Sun Yat-sen University), Xiaobin Chang (Ministry of Education)
ClassificationRecognitionTransformerPrompt EngineeringImage
🎯 What it does: This paper proposes a 'Consistent Prompting (CPrompt)' framework to address the inconsistency between prompt and classifier training and inference in rehearsal-free continual learning.
Consistent3D: Towards Consistent High-Fidelity Text-to-3D Generation with Deterministic Sampling Prior
Zike Wu (Nanyang Technological University), Hanwang Zhang (Nanyang Technological University)
GenerationData SynthesisDiffusion modelNeural Radiance FieldTextPoint CloudOrdinary Differential Equation
🎯 What it does: Proposes the Consistent3D method, which achieves high-fidelity generation from text to 3D using deterministic ODE sampling priors through Consistency Distillation Sampling (CDS) loss.
ConsistNet: Enforcing 3D Consistency for Multi-view Images Diffusion
Jiayu Yang (Tencent), Hongdong Li (Australian National University)
GenerationData SynthesisDiffusion modelImage
🎯 What it does: This paper proposes the ConsistNet module, which inserts lightweight multi-view consistency blocks into the pre-trained Latent Diffusion Model (Zero123-XL) to generate multi-view 3D consistent images from a single input image.
Constrained Layout Generation with Factor Graphs
Mohammed Haroon Dupty (National University of Singapore), Wee Sun Lee (National University of Singapore)
GenerationGraph Neural NetworkGraph
🎯 What it does: Generating floor plan layouts based on spatial constraints using factor graph neural networks.
Construct to Associate: Cooperative Context Learning for Domain Adaptive Point Cloud Segmentation
Guangrui Li (University of Technology Sydney)
SegmentationDomain AdaptationAutonomous DrivingPoint Cloud
🎯 What it does: This paper addresses the domain adaptation problem in point cloud semantic segmentation and proposes a Cooperative Context Learning (CCL) method to achieve adaptation in a setting with fully labeled source domain and unlabeled target domain.
Constructing and Exploring Intermediate Domains in Mixed Domain Semi-supervised Medical Image Segmentation
Qinghe Ma (Nanjing University), Yang Gao (Nanjing University)
SegmentationDomain AdaptationImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: This paper proposes a new mixed-domain semi-supervised medical image segmentation scenario (MiDSS) to address the coexistence of limited annotations and domain shift by constructing an intermediate domain.
Content-Adaptive Non-Local Convolution for Remote Sensing Pansharpening
Yule Duan (University of Electronic Science and Technology of China), Liang-Jian Deng (University of Electronic Science and Technology of China)
RestorationConvolutional Neural NetworkImage
🎯 What it does: This paper proposes Content Adaptive Non-local Convolution (CANConv) and its network CANNet for pansharpening of remote sensing images.