CVPR 2024 Papers — Page 11
IEEE/CVF Conference on Computer Vision and Pattern Recognition · 2716 papers
GES : Generalized Exponential Splatting for Efficient Radiance Field Rendering
Abdullah Hamdi (University of Oxford), Andrea Vedaldi (University of Oxford)
GenerationComputational EfficiencyNeural Radiance FieldGaussian SplattingPoint Cloud
🎯 What it does: A 3D efficient point cloud rendering method based on the generalized exponential function, GES, has been designed and implemented.
GigaPose: Fast and Robust Novel Object Pose Estimation via One Correspondence
Van Nguyen Nguyen (LIGM), Vincent Lepetit (LIGM)
Pose EstimationTransformerContrastive LearningImageBenchmark
🎯 What it does: GigaPose is proposed, a 6D pose estimation method for novel objects that achieves fast and robust coarse pose estimation using limited templates and single image patch correspondences, and can be combined with subsequent refinement methods.
GigaTraj: Predicting Long-term Trajectories of Hundreds of Pedestrians in Gigapixel Complex Scenes
Haozhe Lin (Tsinghua University), Lu Fang (Tsinghua University)
Object TrackingGenerationRecurrent Neural NetworkGenerative Adversarial NetworkVideoMultimodality
🎯 What it does: A GigaTraj dataset was constructed, and various pedestrian trajectory prediction models were evaluated within it.
GLACE: Global Local Accelerated Coordinate Encoding
Fangjinhua Wang (ETH Zurich), Marc Pollefeys (Microsoft)
Pose EstimationRetrievalSimultaneous Localization and MappingImage
🎯 What it does: GLACE proposes a method for achieving high-precision visual localization in large-scale scenes using a single network without the need for 3D point clouds or depth maps.
GLaMM: Pixel Grounding Large Multimodal Model
Hanoona Rasheed (Mohamed bin Zayed University of AI), Fahad S. Khan (Linköping University)
SegmentationGenerationTransformerLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: This paper presents GLaMM, a multimodal large model capable of generating natural language responses in dialogue while simultaneously outputting corresponding pixel-level segmentation masks.
GLID: Pre-training a Generalist Encoder-Decoder Vision Model
Jihao Liu (Chinese University of Hong Kong), Hongsheng Li (Chinese University of Hong Kong)
Object DetectionSegmentationPose EstimationDepth EstimationTransformerImage
🎯 What it does: A general encoder-decoder pre-training method called GLID is proposed, allowing the pre-trained model to be directly fine-tuned for various visual tasks without the need for additional specialized decoders.
GLiDR: Topologically Regularized Graph Generative Network for Sparse LiDAR Point Clouds
Prashant Kumar (Indian Institute of Technology Delhi), Prem Kalra (Indian Institute of Technology Delhi)
SegmentationGenerationAutonomous DrivingGraph Neural NetworkSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Using graph neural networks and 0-dimensional persistent homology regularization for static point completion and dynamic/static segmentation of sparse LiDAR point clouds.
GlitchBench: Can Large Multimodal Models Detect Video Game Glitches?
Mohammad Reza Taesiri (University of Alberta), Anh Nguyen (Auburn University)
Anomaly DetectionTransformerLarge Language ModelVision Language ModelImageMultimodalityBenchmark
🎯 What it does: A multimodal evaluation benchmark called Glitch Bench has been constructed based on real game quality assurance scenarios to detect various anomalies and errors in video games.
Global and Hierarchical Geometry Consistency Priors for Few-shot NeRFs in Indoor Scenes
Xiaotian Sun (Xiamen University), Cheng Wang (Xiamen University)
GenerationData SynthesisDepth EstimationNeural Radiance FieldImage
🎯 What it does: The P NeRF method is proposed, which utilizes the global and hierarchical geometric consistency priors of pre-trained models to improve the NeRF reconstruction quality of indoor scenes from a limited number of views.
Global and Local Prompts Cooperation via Optimal Transport for Federated Learning
Hongxia Li (ShanghaiTech University), Ye Shi (ShanghaiTech University)
OptimizationFederated LearningTransformerPrompt EngineeringVision Language ModelImage
🎯 What it does: This paper proposes FedOTP, a method for tuning visual-language models that simultaneously learns global and local prompts within a federated learning framework, achieving collaboration between the two through unbiased optimal transport.
Global Latent Neural Rendering
Thomas Tanay (Huawei), Matteo Maggioni (Huawei)
GenerationData SynthesisConvolutional Neural NetworkNeural Radiance FieldImage
🎯 What it does: This paper proposes ConvGLR—a new method for global latent space rendering on planar sweeping bodies;
GLOW: Global Layout Aware Attacks on Object Detection
Jun Bao (State Key Laboratory of Blockchain and Data Security), Jun Yu (Hangzhou Dianzi University)
Object DetectionAdversarial AttackImage
🎯 What it does: A global layout-aware adversarial attack generation framework GLOW is proposed for object detectors, which can generate attack plans in scenes with specified or unspecified target objects;
GOAT-Bench: A Benchmark for Multi-Modal Lifelong Navigation
Mukul Khanna (Georgia Institute of Technology), Roozbeh Mottaghi (University of Washington)
Robotic IntelligenceRecurrent Neural NetworkReinforcement LearningVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Proposes the GOAT-Bench benchmark for evaluating robots' lifelong navigation capabilities in open vocabulary and multimodal targets (categories, language descriptions, images).
Going Beyond Multi-Task Dense Prediction with Synergy Embedding Models
Huimin Huang (Zhejiang University), Yefeng Zheng (Guangxi Medical University)
SegmentationDepth EstimationTransformerImage
🎯 What it does: A collaborative embedding model SEM for multi-task dense prediction is proposed, which combines inter-task interaction and an intra-task hierarchical adaptive mechanism to achieve simultaneous optimization of multiple tasks.
GoMAvatar: Efficient Animatable Human Modeling from Monocular Video Using Gaussians-on-Mesh
Jing Wen (University of Illinois), Shenlong Wang (University of Illinois)
GenerationPose EstimationComputational EfficiencyGaussian SplattingVideoMesh
🎯 What it does: This paper presents a real-time, memory-efficient, and animatable single-shot video human model named GoMAvatar, capable of high-quality rendering from any viewpoint and pose.
GoMVS: Geometrically Consistent Cost Aggregation for Multi-View Stereo
Jiang Wu (Northwestern Polytechnical University), Yanning Zhang (Northwestern Polytechnical University)
Depth EstimationConvolutional Neural NetworkImageBenchmark
🎯 What it does: This work proposes a cost aggregation method based on geometric consistency called GoMVS, which projects neighborhood costs into the depth space of the reference pixel using the local plane assumption of adjacent pixels and surface normal vectors, followed by convolutional aggregation.
GoodSAM: Bridging Domain and Capacity Gaps via Segment Anything Model for Distortion-aware Panoramic Semantic Segmentation
Weiming Zhang (Hong Kong University of Science and Technology), Lin Wang (Hong Kong University of Science and Technology)
SegmentationDomain AdaptationKnowledge DistillationTransformerSupervised Fine-TuningImage
🎯 What it does: This paper proposes a knowledge transfer method for unlabeled panoramic images based on the Segment Anything Model (SAM) and Teacher Assistant (TA) to train a lightweight student model for distortion awareness and boundary enhancement in panoramic semantic segmentation.
GOV-NeSF: Generalizable Open-Vocabulary Neural Semantic Fields
Yunsong Wang (National University of Singapore), Gim Hee Lee (National University of Singapore)
SegmentationGenerationNeural Radiance FieldImage
🎯 What it does: A general open vocabulary neural semantic field named GOV-NeSF is proposed for achieving 3D and 2D open vocabulary semantic segmentation and novel view rendering in unseen scenes through multi-view fusion, with training relying solely on 2D RGB images, without the need for depth, semantic labels, or 3D data.
GP-NeRF: Generalized Perception NeRF for Context-Aware 3D Scene Understanding
Hao Li (Northwestern Polytechnical University), Junwei Han (Baidu)
RecognitionSegmentationKnowledge DistillationTransformerNeural Radiance FieldPoint Cloud
🎯 What it does: This paper proposes GP-NeRF, a unified framework that allows NeRF and a 2D segmentation model to work together, using Transformers to jointly construct radiance fields and semantic embedding fields, and enhancing the quality of semantic segmentation through self-distillation.
GPLD3D: Latent Diffusion of 3D Shape Generative Models by Enforcing Geometric and Physical Priors
Yuan Dong (Alibaba Group), Qixing Huang (University of Texas at Austin)
GenerationDiffusion modelPoint Cloud
🎯 What it does: A 3D shape generation model based on latent diffusion, GPLD3D, is proposed, which regularizes the diffusion process by introducing a quality checker for geometric feasibility and physical stability, enhancing the connectivity and structural robustness of the generated shapes.
GPS-Gaussian: Generalizable Pixel-wise 3D Gaussian Splatting for Real-time Human Novel View Synthesis
Shunyuan Zheng (Harbin Institute of Technology), Yebin Liu (Tsinghua University)
GenerationData SynthesisDepth EstimationGaussian SplattingImage
🎯 What it does: A real-time high-resolution human novel viewpoint synthesis method called GPS-Gaussian is proposed, which directly regresses 3D Gaussian points through pixel-level Gaussian parameter mapping, enabling the generation of high-quality viewpoint images without separate optimization.
GPT-4V(ision) is a Human-Aligned Evaluator for Text-to-3D Generation
Tong Wu (Chinese University of Hong Kong), Gordon Wetzstein (Stanford University)
GenerationTransformerLarge Language ModelPrompt EngineeringTextMultimodality
🎯 What it does: A new metric is proposed that utilizes GPT-4V to automatically generate evaluation prompts and conduct multi-dimensional assessments of text-to-3D generation models.
GPT4Point: A Unified Framework for Point-Language Understanding and Generation
Zhangyang Qi (University of Hong Kong), Hengshuang Zhao (University of Hong Kong)
RecognitionGenerationRetrievalTransformerLarge Language ModelDiffusion modelMultimodalityPoint Cloud
🎯 What it does: This paper presents GPT4Point, a unified multimodal framework for 3D point clouds and language, enabling point cloud recognition, description, question answering, and controllable text-to-3D generation.
GraCo: Granularity-Controllable Interactive Segmentation
Yian Zhao (Peking University), Jie Chen (Peking University)
SegmentationTransformerSupervised Fine-TuningImage
🎯 What it does: This work proposes a Granularity-Controllable Interactive Segmentation (GraCo) method that allows users to precisely control the granularity of interactive segmentation through an additional granularity parameter (ranging from 0 to 1), thereby eliminating redundant outputs and addressing spatial ambiguities.
Gradient Alignment for Cross-Domain Face Anti-Spoofing
Binh M. Le (Sungkyunkwan University), Simon S. Woo (Sungkyunkwan University)
RecognitionDomain AdaptationOptimizationConvolutional Neural NetworkImage
🎯 What it does: A gradient alignment learning objective GAC-FAS is proposed for domain generalization in cross-domain face anti-spoofing without additional modules.
Gradient Reweighting: Towards Imbalanced Class-Incremental Learning
Jiangpeng He (Purdue University)
ClassificationKnowledge DistillationImage
🎯 What it does: A continuous incremental learning method for class imbalance is proposed, addressing the imbalance issues both between and within tasks;
Gradient-based Parameter Selection for Efficient Fine-Tuning
Zhi Zhang (University of Amsterdam), Shanghang Zhang (Peking University)
ClassificationSegmentationOptimizationComputational EfficiencyConvolutional Neural NetworkSupervised Fine-TuningContrastive LearningImageBiomedical Data
🎯 What it does: A gradient-based parameter selection method (GPS) is proposed, which only fine-tunes a very small number of parameters in the pre-trained model without adding extra parameters or changing the model structure.
GRAM: Global Reasoning for Multi-Page VQA
Tsachi Blau (Technion), Ron Litman (AWS AI Labs)
TransformerText
🎯 What it does: The GRAM method is proposed, which extends the existing single-page DocVQA model to multi-page documents without the need for additional pre-training, enabling cross-page reasoning while maintaining single-page performance.
GraphDreamer: Compositional 3D Scene Synthesis from Scene Graphs
Gege Gao (Max Planck Institute for Intelligent Systems), Bernhard Schölkopf (Max Planck Institute for Intelligent Systems)
GenerationData SynthesisGraph Neural NetworkDiffusion modelScore-based ModelGraph
🎯 What it does: This paper presents GraphDreamer, a decomposable 3D scene generation framework based on scene graphs, capable of synthesizing high-quality 3D scenes containing multiple separable objects from text or automatically generated scene graphs.
GreedyViG: Dynamic Axial Graph Construction for Efficient Vision GNNs
Mustafa Munir (University of Texas at Austin), Radu Marculescu (University of Texas at Austin)
ClassificationObject DetectionSegmentationComputational EfficiencyKnowledge DistillationConvolutional Neural NetworkGraph Neural NetworkImage
🎯 What it does: A dynamic axial graph construction (DAGC) method is proposed to build an efficient Vision GNN, and a hybrid CNN-GNN structure called GreedyViG is developed based on DAGC to enhance visual task performance while maintaining low parameters and computational power.
Grid Diffusion Models for Text-to-Video Generation
Taegyeong Lee (Ulsan National Institute of Science and Technology), Taehwan Kim (Ulsan National Institute of Science and Technology)
GenerationData SynthesisDiffusion modelVideoText
🎯 What it does: Proposes the Grid Diffusion model, which represents videos as 2×2 or 4×4 grid images, generates key grid images through text conditions, and then uses autoregressive interpolation to obtain the complete video;
Grounded Question-Answering in Long Egocentric Videos
Shangzhe Di (Shanghai Jiao Tong University), Weidi Xie (Shanghai AI Lab)
GenerationRetrievalTransformerLarge Language ModelVideoText
🎯 What it does: This paper proposes a unified model called GroundVQA, which implements spatiotemporal localization and question-answering tasks for long-term egocentric videos.
Grounded Text-to-Image Synthesis with Attention Refocusing
Quynh Phung (University of Maryland), Jia-Bin Huang (University of Maryland)
GenerationData SynthesisTransformerLarge Language ModelDiffusion modelImageText
🎯 What it does: This paper proposes a training-free, attention refocusing framework that utilizes GPT-4 to generate spatial layouts and optimizes cross-attention and self-attention during the diffusion model sampling process, thereby enhancing the controllability of text-to-image generation.
GROUNDHOG: Grounding Large Language Models to Holistic Segmentation
Yichi Zhang (University of Michigan), Joyce Chai (University of Michigan)
SegmentationRetrievalExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodality
🎯 What it does: This paper presents a new multimodal large language model called GROUNDHOG, which can accurately align referable phrases in text to pixel-level segmentation masks, achieving pixel-level visual explanations of natural language.
Grounding and Enhancing Grid-based Models for Neural Fields
Zelin Zhao (Shanghai Jiao Tong University), Junchi Yan (Shanghai Jiao Tong University)
Super ResolutionNeural Radiance FieldImagePoint Cloud
🎯 What it does: A theoretical framework for a grid-based neural field model is proposed, and a new MulFAGrid model is designed.
Grounding Everything: Emerging Localization Properties in Vision-Language Transformers
Walid Bousselham (University of Bonn), Hilde Kuehne (Goethe University Frankfurt)
Object DetectionSegmentationTransformerVision Language ModelContrastive LearningImageMultimodality
🎯 What it does: Based on a pre-trained visual-language transformer, a training-free 'Grounding Everything Module' (GEM) is proposed, which achieves object localization and segmentation of open vocabulary through self-self attention.
GroupContrast: Semantic-aware Self-supervised Representation Learning for 3D Understanding
Chengyao Wang (Chinese University of Hong Kong), Jiaya Jia (Chinese University of Hong Kong)
Object DetectionSegmentationKnowledge DistillationRepresentation LearningContrastive LearningPoint Cloud
🎯 What it does: A self-supervised 3D point cloud representation learning framework called GroupContrast is proposed, which combines segment grouping and semantic-aware contrastive learning to enhance the representation consistency of semantically similar points in point clouds.
Groupwise Query Specialization and Quality-Aware Multi-Assignment for Transformer-based Visual Relationship Detection
Jongha Kim (Korea University), Hyunwoo J. Kim (Korea University)
RecognitionObject DetectionTransformerImage
🎯 What it does: This paper proposes SpeaQ, a specialized and quality-aware multi-assignment label allocation method for Transformer-based visual relation detection.
GS-IR: 3D Gaussian Splatting for Inverse Rendering
Zhihao Liang (South China University of Technology), Kui Jia (Chinese University of Hong Kong)
RestorationGenerationNeural Radiance FieldGaussian SplattingImage
🎯 What it does: Using 3D Gaussian Splatting for inverse rendering to recover the geometry, materials, and lighting of a scene, generating high-quality new view renderings and relighting from multi-view images.
GS-SLAM: Dense Visual SLAM with 3D Gaussian Splatting
Chi Yan (Shanghai AI Laboratory), Xuelong Li (TeleAI, China Telecom Corporation Limited)
Pose EstimationOptimizationComputational EfficiencyGaussian SplattingSimultaneous Localization and MappingPoint Cloud
🎯 What it does: This paper presents GS-SLAM, a real-time RGB-D dense SLAM system based on 3D Gaussian decomposition, achieving efficient map construction and camera pose tracking.
GSNeRF: Generalizable Semantic Neural Radiance Fields with Enhanced 3D Scene Understanding
Zi-Ting Chou (National Taiwan University), Yu-Chiang Frank Wang (NVIDIA)
SegmentationGenerationDepth EstimationConvolutional Neural NetworkNeural Radiance FieldImage
🎯 What it does: This paper proposes a generalizable Semantic Neural Radiance Field (GSNeRF) that can generate synthetic view images and corresponding semantic segmentation maps of unseen scenes with only multi-view images and camera poses provided.
GSVA: Generalized Segmentation via Multimodal Large Language Models
Zhuofan Xia (Tsinghua University), Gao Huang (Tsinghua University)
Object DetectionSegmentationTransformerLarge Language ModelPrompt EngineeringImageMultimodality
🎯 What it does: A multimodal large language model framework named GSVA is proposed, capable of simultaneously handling multi-object segmentation and empty object recognition in the General Reference Expression Segmentation (GRES) task.
Guess The Unseen: Dynamic 3D Scene Reconstruction from Partial 2D Glimpses
Inhee Lee (Seoul National University), Hanbyul Joo (Seoul National University)
GenerationData SynthesisPose EstimationDiffusion modelGaussian SplattingVideo
🎯 What it does: Using sparse 2D observations from monocular videos, we reconstruct a four-dimensional 3D scene that includes a static background and multiple humans, supporting rendering and editing from arbitrary viewpoints and poses.
Guided Slot Attention for Unsupervised Video Object Segmentation
Minhyeok Lee (Yonsei University), Sangyoun Lee (Yonsei University)
Object DetectionSegmentationTransformerOptical FlowVideo
🎯 What it does: This paper proposes an unsupervised video object segmentation network called GSA-Net based on Guided Slot Attention, which can accurately segment foreground objects in complex backgrounds and multi-object scenes.
H-ViT: A Hierarchical Vision Transformer for Deformable Image Registration
Morteza Ghahremani (Technical University of Munich), Christian Wachinger (Technical University of Munich)
TransformerImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: This paper proposes a Hierarchical Visual Transformer (H-ViT) for deformable image registration, which encodes the deformation field through a dual attention mechanism.
Habitat Synthetic Scenes Dataset (HSSD-200): An Analysis of 3D Scene Scale and Realism Tradeoffs for ObjectGoal Navigation
Mukul Khanna (Georgia Institute of Technology), Manolis Savva (Simon Fraser University)
Robotic IntelligenceReinforcement LearningImage
🎯 What it does: A high-quality, artificially designed 3D indoor scene dataset HSSD-200 was constructed, and it was used to evaluate the generalization ability of navigation agents in real environments.
HalluciDoctor: Mitigating Hallucinatory Toxicity in Visual Instruction Data
Qifan Yu (Zhejiang University), Yueting Zhuang (Zhejiang University)
GenerationData-Centric LearningTransformerLarge Language ModelVision Language ModelImageMultimodality
🎯 What it does: This paper proposes a cross-checking framework named HalluciDoctor, which is used to automatically detect and eliminate hallucination toxicity in machine-generated visual instruction data, and enhance the robustness of multimodal large language models through adversarial visual instruction augmentation.
Hallucination Augmented Contrastive Learning for Multimodal Large Language Model
Chaoya Jiang (National Engineering Research Center for Software Engineering Peking University), Shikun Zhang (National Engineering Research Center for Software Engineering Peking University)
Representation LearningTransformerLarge Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: This paper proposes a new training method for multimodal large language models—Hallucination Augmented Contrastive Learning (HACL), which enhances the alignment of visual and language representations and suppresses hallucination generation by incorporating hallucination descriptions generated by GPT-4 as hard negative samples in contrastive learning between vision and text.
HallusionBench: An Advanced Diagnostic Suite for Entangled Language Hallucination and Visual Illusion in Large Vision-Language Models
Tianrui Guan (University of Maryland), Tianyi Zhou (University of Maryland)
RecognitionData SynthesisTransformerLarge Language ModelVision Language ModelImageMultimodalityBenchmark
🎯 What it does: A benchmark called HallusionBench is proposed to systematically evaluate large visual language models in terms of language hallucinations and visual illusions in image context reasoning.
HandBooster: Boosting 3D Hand-Mesh Reconstruction by Conditional Synthesis and Sampling of Hand-Object Interactions
Hao Xu (Chinese University of Hong Kong), Chi-Wing Fu (Chinese University of Hong Kong)
GenerationData SynthesisPose EstimationDiffusion modelImageMesh
🎯 What it does: Utilize conditional diffusion models to generate realistic hand-object interaction images, and achieve controllable synthesis by constructing content-aware conditions (normal maps, texture maps, and hand pose quaternions);
HandDiff: 3D Hand Pose Estimation with Diffusion on Image-Point Cloud
Wencan Cheng (Sungkyunkwan University), Jong Hwan Ko (Sungkyunkwan University)
Pose EstimationGraph Neural NetworkDiffusion modelImagePoint Cloud
🎯 What it does: This paper proposes a 3D hand pose estimation method called HandDiff based on diffusion models, which utilizes the joint conditions of depth images and point clouds for iterative denoising.
HanDiffuser: Text-to-Image Generation With Realistic Hand Appearances
Supreeth Narasimhaswamy (Stony Brook University), Minh Hoai (Adobe Research)
GenerationData SynthesisDiffusion modelImageTextMultimodality
🎯 What it does: An end-to-end text-to-image generation model called HanDiffuser is proposed, capable of generating high-quality images of hands.
HardMo: A Large-Scale Hardcase Dataset for Motion Capture
Jiaqi Liao (Beijing University of Posts and Telecommunications), Junran Peng (University of Science and Technology Beijing)
Pose EstimationOptimizationSupervised Fine-TuningVideo
🎯 What it does: A large-scale monocular motion capture dataset of dance and martial arts actions, named HardMo, has been constructed, from which challenging samples for hands and feet have been mined and optimized, resulting in the HardMo-Hand and HardMo-Foot subsets.
HarmonyView: Harmonizing Consistency and Diversity in One-Image-to-3D
Sangmin Woo (KAIST), Changick Kim (KAIST)
GenerationData SynthesisDiffusion modelNeural Radiance FieldImage
🎯 What it does: Developed HarmonyView, a diffusion sampling technique that simultaneously considers consistency and diversity in single-image to 3D generation.
Harnessing Large Language Models for Training-free Video Anomaly Detection
Luca Zanella (University of Trento), Elisa Ricci (University of Trento)
Anomaly DetectionTransformerLarge Language ModelPrompt EngineeringVision Language ModelVideoMultimodality
🎯 What it does: A training-free video anomaly detection method called LAVAD is proposed, which utilizes a pre-trained vision-language model to generate video frame captions, then scores anomalies using a large language model, and combines cross-modal similarity for caption cleaning and score refinement.
Harnessing Meta-Learning for Improving Full-Frame Video Stabilization
Muhammad Kashif Ali (Hanyang University), Tae Hyun Kim (Hanyang University)
RestorationMeta LearningOptical FlowVideo
🎯 What it does: A meta-learning-based adaptive method for testing is proposed, which can quickly adjust the parameters of a full-frame pixel-level video stabilization model, thereby improving the stability and quality of a single video.
Harnessing the Power of MLLMs for Transferable Text-to-Image Person ReID
Wentan Tan, Dapeng Tao (Yunnan University)
RetrievalTransformerLarge Language ModelPrompt EngineeringImageTextMultimodality
🎯 What it does: This paper constructs a large-scale transferable human retrieval dataset by automatically generating text descriptions using a multimodal large language model and combining noise suppression techniques, significantly improving cross-domain text retrieval performance.
HashPoint: Accelerated Point Searching and Sampling for Neural Rendering
Jiahao Ma (Australian National University), Chuong Nguyen (CSIRO Data61)
Autonomous DrivingComputational EfficiencyNeural Radiance FieldSimultaneous Localization and MappingPoint Cloud
🎯 What it does: The HashPoint method is proposed, which combines rasterization and ray tracing techniques to accelerate point cloud search through a hash table and adaptively sample near the primary surface, significantly improving the speed and efficiency of neural rendering.
HAVE-FUN: Human Avatar Reconstruction from Few-Shot Unconstrained Images
Xihe Yang (Xiaobing AI), Baoyuan Wang (Xiaobing AI)
GenerationPose EstimationDiffusion modelScore-based ModelImageMesh
🎯 What it does: A framework named HaveFun is proposed, which can reconstruct animatable 3D models of the human body and hands using only 2-8 casually taken photos.
HDQMF: Holographic Feature Decomposition Using Quantum Algorithms
Prathyush Prasanth Poduval (University of California), Mohsen Imani (University of California)
🎯 What it does: This paper proposes a hyper-dimensional vector decomposition method based on quantum computing (HDQMF) to address the memory decomposition problem in HDC.
HDRFlow: Real-Time HDR Video Reconstruction with Large Motions
Gangwei Xu (Huazhong University of Science and Technology), Xin Yang (Huazhong University of Science and Technology)
RestorationOptical FlowVideo
🎯 What it does: A real-time HDR video reconstruction method called HDRFlow is proposed, specifically designed for alternating exposure sequences, addressing ghosting caused by large motion and exposure inconsistencies.
HEAL-SWIN: A Vision Transformer On The Sphere
Oscar Carlsson (Chalmers University of Technology), Daniel Persson (Chalmers University of Technology)
SegmentationDepth EstimationAutonomous DrivingTransformerImage
🎯 What it does: A Transformer model named HEAL-SWIN is proposed, which maps high-resolution fisheye images onto a sphere using HEALPix equal-area equidistant pixel grids, and performs tasks such as semantic segmentation, depth estimation, and classification based on this mapping.
Hearing Anything Anywhere
Mason Long Wang (Stanford University), Jiajun Wu (Stanford University)
Diffusion modelAudio
🎯 What it does: This paper presents the DIFFRIR framework, which synthesizes realistic spatial audio at any room position using a small number of RIR measurements and simple geometric models.
HHMR: Holistic Hand Mesh Recovery by Enhancing the Multimodal Controllability of Graph Diffusion Models
Mengcheng Li (Tsinghua University), Yebin Liu (Tsinghua University)
RestorationGenerationPose EstimationGraph Neural NetworkDiffusion modelImageMultimodalityMesh
🎯 What it does: A full-process hand mesh recovery framework (HHMR) based on a graph diffusion model is proposed, which can accomplish hand mesh generation, completion, single-image recovery, and 2D skeleton fitting within the same model.
Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds
Tianrui Lou (Sun Yat-Sen University), Xiaochun Cao (Sun Yat-Sen University)
ClassificationAdversarial AttackGraph Neural NetworkGaussian SplattingPoint Cloud
🎯 What it does: We propose HiT-ADV, a 3D point cloud adversarial attack method based on shape deformation, which can generate effective perturbations for the model without producing obvious outliers and is almost imperceptible to the human eye.
Hierarchical Correlation Clustering and Tree Preserving Embedding
Morteza Haghir Chehreghani (Chalmers University of Technology), Mostafa Haghir Chehreghani (Amirkabir University of Technology)
OptimizationComputational EfficiencyTabular
🎯 What it does: A hierarchical clustering method called HCC is proposed, which can handle both positive and negative similarities. Based on its tree structure, a tree-preserving embedding is designed to generate features that can be used for subsequent clustering (such as GMM). The combination of minimax distance and related clustering is also studied, significantly reducing computational complexity and focusing on clusters with varying shapes.
Hierarchical Diffusion Policy for Kinematics-Aware Multi-Task Robotic Manipulation
Xiao Ma (Dyson Robot Learning Lab), Stephen James (Dyson Robot Learning Lab)
Robotic IntelligenceReinforcement LearningDiffusion modelPoint Cloud
🎯 What it does: A hierarchical robot control framework is proposed—using Next Best Pose (NBP) prediction at the high level and a diffusion model of differentiable robot kinematics (RK-Diffuser) to generate joint trajectories at the low level, achieving precise control for multi-task and language-guided scenarios.
Hierarchical Histogram Threshold Segmentation - Auto-terminating High-detail Oversegmentation
Thomas V. Chang (Nuremberg Institute of Technology), Bartosz von Rymon Lipinski (Nuremberg Institute of Technology)
SegmentationImage
🎯 What it does: High-quality superpixels are obtained through adaptive hierarchical histogram threshold segmentation, without the need for seed initialization or manual termination parameters.
Hierarchical Intra-modal Correlation Learning for Label-free 3D Semantic Segmentation
Xin Kang (University of Science and Technology of China), Yan Lu (Microsoft Research Asia)
SegmentationAutonomous DrivingKnowledge DistillationTransformerVision Language ModelPoint Cloud
🎯 What it does: A hierarchical internal modal association learning framework is proposed for unsupervised 3D semantic segmentation, primarily improving pseudo-labels and enhancing the compactness of the feature space by modeling the visual and geometric associations between points within the same geometric set, the same scene, and different scenes.
Hierarchical Patch Diffusion Models for High-Resolution Video Generation
Ivan Skorokhodov (Snap Inc), Sergey Tulyakov
GenerationData SynthesisDiffusion modelVideoText
🎯 What it does: Designed and trained the Hierarchical Patch Diffusion Model (HPDM) to achieve end-to-end high-resolution video generation.
Hierarchical Spatio-temporal Decoupling for Text-to-Video Generation
Zhiwu Qing (Huazhong University of Science and Technology), Nong Sang (Huazhong University of Science and Technology)
GenerationDiffusion modelVideoText
🎯 What it does: This paper presents HiGen, a text-to-video generation method based on diffusion models.
HiFi4G: High-Fidelity Human Performance Rendering via Compact Gaussian Splatting
Yuheng Jiang (ShanghaiTech University), Lan Xu (ShanghaiTech University)
GenerationCompressionOptimizationGaussian SplattingVideo
🎯 What it does: We propose HiFi4G, an explicit 4D human performance rendering framework based on compressed efficient Gaussian Splatting, which achieves real-time high-quality rendering by combining non-rigid tracking.
HIG: Hierarchical Interlacement Graph Approach to Scene Graph Generation in Video Understanding
Trong-Thuan Nguyen (University of Arkansas), Khoa Luu (University of Arkansas)
Object DetectionGenerationGraph Neural NetworkVideoGraph
🎯 What it does: This paper proposes a visual interaction understanding task and constructs the ASPIRe dataset, which includes five types of interaction predicates (Appearance, Situation, Position, Interaction, Relation), and designs a Hierarchical Interlacement Graph (HIG) model to generate video scene graphs.
High-fidelity Person-centric Subject-to-Image Synthesis
Yibin Wang (Fudan University), Cheng Jin (Fudan University)
GenerationData SynthesisDiffusion modelImage
🎯 What it does: Introduces the Face-diffuser framework, which uses two specialized diffusion models (TDM and SDM) to achieve high-fidelity human generation through a three-stage sampling process.
High-Quality Facial Geometry and Appearance Capture at Home
Yuxuan Han (Tsinghua University), Feng Xu (Tsinghua University)
Neural Radiance FieldImageVideoMesh
🎯 What it does: Using a mobile phone camera and flashlight to capture a single sequence in a dimly lit room, a low-cost and user-friendly complete facial geometry and material capture system is constructed, which can export results as 3D assets compatible with common CG software such as Blender.
Higher-order Relational Reasoning for Pedestrian Trajectory Prediction
Sungjune Kim (Korea University), Sangpil Kim (Korea University)
Graph Neural NetworkGraphTime Series
🎯 What it does: A module named HighGraph has been researched and implemented to model higher-order social relationships in pedestrian trajectory prediction.
HiKER-SGG: Hierarchical Knowledge Enhanced Robust Scene Graph Generation
Ce Zhang (Carnegie Mellon University), Yaqi Xie (Carnegie Mellon University)
Object DetectionGenerationGraph Neural NetworkImageBenchmark
🎯 What it does: Proposes the HiKER‑SGG method, which utilizes hierarchical knowledge graphs and layered reasoning to achieve robust scene graph generation, and constructs the VG‑C (20 types of distortion) benchmark.
HiLo: Detailed and Robust 3D Clothed Human Reconstruction with High-and Low-Frequency Information of Parametric Models
Yifan Yang (South China University of Technology), Mingkui Tan (South China University of Technology)
GenerationData SynthesisPose EstimationImagePoint CloudMesh
🎯 What it does: This paper proposes a method for generating 3D clothed human models from single-view color images using a parametric human model that utilizes high-frequency and low-frequency information, called HiLo.
HIMap: HybrId Representation Learning for End-to-end Vectorized HD Map Construction
Yi Zhou (Samsung Research and Development Institute China Beijing), ByungIn Yoo (Samsung Research and Development Institute China Beijing)
Autonomous DrivingRepresentation LearningTransformerImagePoint Cloud
🎯 What it does: Construct an end-to-end vectorized high-precision map, predicting the category and point coordinates of map elements.
HINTED: Hard Instance Enhanced Detector with Mixed-Density Feature Fusion for Sparsely-Supervised 3D Object Detection
Qiming Xia (Xiamen University), Cheng Wang (Xiamen University)
Object DetectionAutonomous DrivingPoint Cloud
🎯 What it does: This paper proposes a Hard Instance Enhanced Detector (HINTED) for sparse supervised 3D object detection, which significantly improves detection performance in sparsely labeled scenarios through a Self-Boosting Teacher (SBT) and Mixed Density Student (MDS) framework.
HiPose: Hierarchical Binary Surface Encoding and Correspondence Pruning for RGB-D 6DoF Object Pose Estimation
Yongliang Lin (Zhejiang University), Yu Zhang (Zhejiang University)
Pose EstimationConvolutional Neural NetworkImage
🎯 What it does: A rendering-free, RANSAC-free 6DoF pose estimation method based on RGB-D called HiPose is proposed.
HIPTrack: Visual Tracking with Historical Prompts
Wenrui Cai (Beihang University), Yunhong Wang (Beihang University)
Object TrackingTransformerVideo
🎯 What it does: This paper proposes the Historical Prompt Network and the HIPTrack tracker based on this network, which generates high-quality prompts by integrating precise foreground masks and visual features of historical targets within the traditional Siamese tracking framework, thereby improving tracking accuracy.
HIR-Diff: Unsupervised Hyperspectral Image Restoration Via Improved Diffusion Models
Li Pang (Xi'an Jiaotong University), Xiangyong Cao (Xi'an Jiaotong University)
RestorationSuper ResolutionDiffusion modelImage
🎯 What it does: A framework for unsupervised hyperspectral image restoration based on a pre-trained diffusion model (HIR-Diff) is proposed. It decomposes hyperspectral images into low-dimensional images and coefficient matrices through low-rank decomposition. The coefficient matrix is first estimated using SVD+RRQR, and then the low-dimensional image is restored using an improved diffusion model with a total variation (TV) regularization guidance function, ultimately reconstructing a clear hyperspectral image.
HIT: Estimating Internal Human Implicit Tissues from the Body Surface
Marilyn Keller (Max Planck Institute for Intelligent Systems), Sergi Pujades (Helmholtz Center Munich)
SegmentationImageBiomedical DataMagnetic Resonance ImagingBenchmark
🎯 What it does: Using surface information (SMPL model) to learn an implicit volume model (HIT) that can predict the position and volume of three important internal tissues in the human body: subcutaneous fat, lean tissue (muscle and organs), and long bones.
HIVE: Harnessing Human Feedback for Instructional Visual Editing
Shu Zhang (Salesforce AI Research), Ran Xu (Stanford University)
GenerationData SynthesisReinforcement LearningDiffusion modelImage
🎯 What it does: Proposes the HIVE framework, which integrates human feedback into the diffusion model for instruction-based image editing;
HMD-Poser: On-Device Real-time Human Motion Tracking from Scalable Sparse Observations
Peng Dai (ByteDance), Zeming Li (ByteDance)
Pose EstimationRecurrent Neural NetworkTransformerTime Series
🎯 What it does: This paper proposes HMD-Poser, which utilizes HMD and scalable IMU sensors to achieve real-time full-body motion tracking, enabling online inference on consumer-grade HMDs.
HOI-M^3: Capture Multiple Humans and Objects Interaction within Contextual Environment
Juze Zhang (ShanghaiTech University), Jingya Wang (ShanghaiTech University)
GenerationPose EstimationDiffusion modelVideoPoint CloudBenchmark
🎯 What it does: The first multi-person multi-object 3D interaction dataset HOI-M 3 has been constructed, and based on this dataset, two benchmark tasks of monocular multi-body-object interaction capture and multi-interaction generation have been proposed.
HOIAnimator: Generating Text-prompt Human-object Animations using Novel Perceptive Diffusion Models
Wenfeng Song (Beijing Information Science and Technology University), Hong Qin (China University of Petroleum)
GenerationTransformerDiffusion modelVideo
🎯 What it does: Generate 3D human-object interaction animations through text prompts, constructing complete animation sequences.
HOIDiffusion: Generating Realistic 3D Hand-Object Interaction Data
Mengqi Zhang (University of California San Diego), Xiaolong Wang (University of California San Diego)
GenerationData SynthesisPose EstimationDiffusion modelImage
🎯 What it does: Proposes the HOIDiffusion model for generating images with realistic 3D hand-object interactions and corresponding 3D annotations, supporting decoupled control of structure and appearance;
HOISDF: Constraining 3D Hand-Object Pose Estimation with Global Signed Distance Fields
Haozhe Qi (Ecole Polytechnique Federale de Lausanne), Alexander Mathis (Ecole Polytechnique Federale de Lausanne)
Pose EstimationConvolutional Neural NetworkTransformerImage
🎯 What it does: This paper proposes an end-to-end network (HOISDF) that utilizes a global implicit Signed Distance Field (SDF) to guide 3D pose estimation of hand-object interactions.
HOIST-Former: Hand-held Objects Identification Segmentation and Tracking in the Wild
Supreeth Narasimhaswamy (Stony Brook University), Minh Hoai (VinAI Research)
Object DetectionObject TrackingSegmentationTransformerVideo
🎯 What it does: Proposes HOIST-Former, a Transformer architecture for recognizing, segmenting, and tracking handheld objects.
HOLD: Category-agnostic 3D Reconstruction of Interacting Hands and Objects from Video
Zicong Fan (ETH Zurich), Otmar Hilliges (ETH Zurich)
Object DetectionPose EstimationNeural Radiance FieldVideo
🎯 What it does: A method named HOLD is proposed, which can simultaneously reconstruct the 3D shape and pose of the hand and objects from a monocular video without relying on pre-scanned object templates or category constraints.
Holistic Autonomous Driving Understanding by Bird's-Eye-View Injected Multi-Modal Large Models
Xinpeng Ding (Hong Kong University of Science and Technology), Xiaomeng Li (Hong Kong University of Science and Technology)
Autonomous DrivingTransformerLarge Language ModelVision Language ModelVideoTextMultimodality
🎯 What it does: Proposed the NuInstruct dataset and the BEV-InMLLM model for language-driven multimodal autonomous driving understanding.
Holistic Features are almost Sufficient for Text-to-Video Retrieval
Kaibin Tian (Renmin University of China), Xirong Li (Renmin University of China)
RetrievalKnowledge DistillationTransformerContrastive LearningVideoText
🎯 What it does: This paper proposes the TeachCLIP framework, which incorporates an Attentional frame-Feature Aggregation (AFA) block into CLIP4Clip, enabling the distillation of fine-grained cross-modal knowledge from a heavy teacher model to a lightweight student model for efficient text-video retrieval.
Holo-Relighting: Controllable Volumetric Portrait Relighting from a Single Image
Yiqun Mei (Johns Hopkins University), Vishal M. Patel (Johns Hopkins University)
GenerationData SynthesisConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: A controllable volume portrait relighting method Holo-Relighting is proposed for single image, achieving free viewpoint and lighting re-rendering.
Holodeck: Language Guided Generation of 3D Embodied AI Environments
Yue Yang (University of Pennsylvania), Christopher Clark (University of Pennsylvania)
GenerationRobotic IntelligenceTransformerLarge Language ModelPrompt EngineeringPoint CloudMesh
🎯 What it does: This paper presents HOLODECK, a system that utilizes large language models to automatically generate diverse and customizable interactive 3D Embodied AI environments.
Holoported Characters: Real-time Free-viewpoint Rendering of Humans from Sparse RGB Cameras
Ashwath Shetty (Max Planck Institute for Informatics), Christian Theobalt (Max Planck Institute for Informatics)
GenerationPose EstimationSuper ResolutionConvolutional Neural NetworkGraph Neural NetworkImageVideo
🎯 What it does: A three-stage real-time 4K resolution sparse RGB camera + 3D skeletal pose input method for human free viewpoint rendering is proposed.
HoloVIC: Large-scale Dataset and Benchmark for Multi-Sensor Holographic Intersection and Vehicle-Infrastructure Cooperative
Cong Ma (SenseAuto Research), Wei Wu (Tsinghua University)
Object DetectionObject TrackingAutonomous DrivingVideoPoint CloudBenchmark
🎯 What it does: This paper constructs a multi-sensor holographic intersection dataset HoloVIC, which includes cameras, LiDAR, and fisheye cameras, providing over 100k synchronized frames and 11.47M 3D boxes with cross-device and temporal ID annotations, and proposes 5 tasks along with baseline models.
HomoFormer: Homogenized Transformer for Image Shadow Removal
Jie Xiao (University of Science and Technology of China), Zheng-Jun Zha (University of Science and Technology of China)
RestorationTransformerImage
🎯 What it does: This paper studies a local window Transformer (HomoFormer) that achieves spatial homogenization using random shuffling and inverse shuffling for high-resolution image shadow removal.
Honeybee: Locality-enhanced Projector for Multimodal LLM
Junbum Cha (Kakao Brain), Byungseok Roh (Kakao Brain)
Convolutional Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Designed local enhancement projectors (C-Abstractor, D-Abstractor) and proposed a complete training and instruction recipe for multimodal large language models, constructing a multimodal visual instruction tuning system called Honeybee.