cs.CV（2025-11-24）

📊 共 38 篇论文 | 🔗 7 篇有代码

🎯 兴趣领域导航

支柱三：空间感知 (Perception & SLAM) (21 🔗5) 支柱二：RL算法与架构 (RL & Architecture) (5) 支柱一：机器人控制 (Robot Control) (5 🔗2) 支柱五：交互与反应 (Interaction & Reaction) (3) 支柱四：生成式动作 (Generative Motion) (2) 支柱七：动作重定向 (Motion Retargeting) (1) 支柱六：视频提取与匹配 (Video Extraction & Matching) (1)

🔬 支柱三：空间感知 (Perception & SLAM) (21 篇)

#题目一句话要点标签🔗⭐
1 DensifyBeforehand: LiDAR-assisted Content-aware Densification for Efficient and Quality 3D Gaussian Splatting 提出LiDAR辅助的内容感知稠密化方法，提升3D高斯溅射效率与质量 depth estimation monocular depth 3D gaussian splatting
2 Neural Texture Splatting: Expressive 3D Gaussian Splatting for View Synthesis, Geometry, and Dynamic Reconstruction 提出神经纹理溅射（NTS），提升3D高斯溅射在视图合成、几何及动态重建任务上的性能。 3D gaussian splatting 3DGS gaussian splatting
3 IDSplat: Instance-Decomposed 3D Gaussian Splatting for Driving Scenes IDSplat：面向自动驾驶场景的实例分解3D高斯溅射重建 3D gaussian splatting gaussian splatting
4 NVGS: Neural Visibility for Occlusion Culling in 3D Gaussian Splatting 提出基于神经可见性的3D高斯溅射遮挡剔除方法，提升复杂场景渲染效率。 3D gaussian splatting gaussian splatting
5 MapRF: Weakly Supervised Online HD Map Construction via NeRF-Guided Self-Training MapRF：基于NeRF引导自训练的弱监督在线高清地图构建 NeRF neural radiance
6 Sphinx: Efficiently Serving Novel View Synthesis using Regression-Guided Selective Refinement Sphinx：提出一种基于回归引导选择性优化的高效新视角合成框架 novel view synthesis navigation
7 MetroGS: Efficient and Stable Reconstruction of Geometrically Accurate High-Fidelity Large-Scale Scenes MetroGS：高效稳定地重建几何精确的高保真大规模场景 3D gaussian splatting gaussian splatting scene reconstruction
8 Graph-based 3D Human Pose Estimation using WiFi Signals 提出GraphPose-Fi，利用WiFi信号和图神经网络进行3D人体姿态估计 pose estimation ✅
9 Robust Long-term Test-Time Adaptation for 3D Human Pose Estimation through Motion Discretization 提出基于运动离散化的鲁棒长期测试时自适应3D人体姿态估计方法 pose estimation
10 Proxy-Free Gaussian Splats Deformation with Splat-Based Surface Estimation 提出无代理高斯斑点变形方法以解决表面信息捕捉不足问题 NeRF point cloud ✅
11 The Determinant Ratio Matrix Approach to Solving 3D Matching and 2D Orthographic Projection Alignment Tasks 提出基于行列式比率矩阵（DRaM）的EnP和OnP问题求解方法 pose estimation 3D pose estimation
12 Prune-Then-Plan: Step-Level Calibration for Stable Frontier Exploration in Embodied Question Answering Prune-Then-Plan：通过步级校准实现具身问答中稳定的边界探索 navigation
13 A Storage-Efficient Feature for 3D Concrete Defect Segmentation to Replace Normal Vector 提出基于相对角度的3D混凝土缺陷分割特征，实现存储效率提升。 point cloud
14 IndEgo: A Dataset of Industrial Scenarios and Collaborative Work for Egocentric Assistants IndEgo：用于第一人称视角工业助手协作任务的多模态数据集 point cloud ✅
15 From Pixels to Posts: Retrieval-Augmented Fashion Captioning and Hashtag Generation 提出检索增强的时尚描述与标签生成框架，提升属性保真度和领域泛化性。 localization
16 DiffSeg30k: A Multi-Turn Diffusion Editing Benchmark for Localized AIGC Detection DiffSeg30k：用于AIGC精细化检测的多轮扩散编辑基准数据集 localization ✅
17 Granular Computing-driven SAM: From Coarse-to-Fine Guidance for Prompt-Free Segmentation 提出基于粒计算的Grc-SAM，实现无提示图像分割的粗到细精度提升。 localization
18 Perceptual Taxonomy: Evaluating and Guiding Hierarchical Scene Reasoning in Vision-Language Models 提出感知分类法，用于评估和指导视觉-语言模型中的分层场景推理 scene understanding
19 TPG-INR: Target Prior-Guided Implicit 3D CT Reconstruction for Enhanced Sparse-view Imaging 提出TPG-INR以解决超稀视图下3D CT重建精度不足问题 NeRF ✅
20 PartDiffuser: Part-wise 3D Mesh Generation via Discrete Diffusion PartDiffuser：通过离散扩散实现分部件的三维网格生成 point cloud
21 Understanding Task Transfer in Vision-Language Models 提出Perfection Gap Factor，系统研究视觉-语言模型中的任务迁移现象 depth estimation

🔬 支柱二：RL算法与架构 (RL & Architecture) (5 篇)

#题目一句话要点标签🔗⭐
22 FineXtrol: Controllable Motion Generation via Fine-Grained Text FineXtrol：通过细粒度文本控制实现可控的运动生成 contrastive learning text-driven motion motion generation
23 ReEXplore: Improving MLLMs for Embodied Exploration with Contextualized Retrospective Experience Replay ReEXplore：利用情境化回顾经验回放改进MLLM在具身探索中的性能 reinforcement learning imitation learning navigation
24 VideoChat-M1: Collaborative Policy Planning for Video Understanding via Multi-Agent Reinforcement Learning 提出VideoChat-M1，通过多智能体强化学习实现视频理解的协同策略规划。 reinforcement learning
25 Fewer Tokens, Greater Scaling: Self-Adaptive Visual Bases for Efficient and Expansive Representation Learning 提出自适应视觉基，减少视觉Token数量，提升视觉表征学习的效率和可扩展性 representation learning
26 MagicWorld: Interactive Geometry-driven Video World Exploration MagicWorld：提出几何引导的交互式视频世界探索模型，提升场景稳定性和连续性。 world model point cloud

🔬 支柱一：机器人控制 (Robot Control) (5 篇)

#题目一句话要点标签🔗⭐
27 AuViRe: Audio-visual Speech Representation Reconstruction for Deepfake Temporal Localization 提出AuViRe，通过音视频语音表征重建实现Deepfake视频的时间定位 manipulation localization ✅
28 UMCL: Unimodal-generated Multimodal Contrastive Learning for Cross-compression-rate Deepfake Detection 提出UMCL框架，通过单模态生成多模态对比学习，解决跨压缩率深度伪造检测难题。 manipulation contrastive learning
29 GuideFlow: Constraint-Guided Flow Matching for Planning in End-to-End Autonomous Driving GuideFlow：一种约束引导的Flow Matching方法，用于端到端自动驾驶规划。 manipulation flow matching ✅
30 LAA3D: A Benchmark of Detecting and Tracking Low-Altitude Aircraft in 3D Space LAA3D：构建低空飞行器三维感知基准数据集与单目3D检测基线。 sim-to-real pose estimation localization
31 VDC-Agent: When Video Detailed Captioners Evolve Themselves via Agentic Self-Reflection VDC-Agent：通过Agent自反思进化视频详细描述模型，无需人工标注和大型教师模型。 running

🔬 支柱五：交互与反应 (Interaction & Reaction) (3 篇)

#题目一句话要点标签🔗⭐
32 Mitigating Long-Tail Bias in HOI Detection via Adaptive Diversity Cache 提出自适应多样性缓存模块，无需额外训练即可缓解HOI检测中的长尾偏差。 human-object interaction HOI
33 SyncMV4D: Synchronized Multi-view Joint Diffusion of Appearance and Motion for Hand-Object Interaction Synthesis SyncMV4D：同步多视角联合扩散生成手-物交互视频与4D运动 HOI
34 Peregrine: One-Shot Fine-Tuning for FHE Inference of General Deep CNNs Peregrine：用于通用深度CNN的FHE推理的单次微调方法 OMOMO

🔬 支柱四：生成式动作 (Generative Motion) (2 篇)

#题目一句话要点标签🔗⭐
35 ReAlign: Text-to-Motion Generation via Step-Aware Reward-Guided Alignment 提出ReAlign，通过步进式奖励引导对齐实现高质量文本到动作生成 text-to-motion motion generation
36 Rethinking Garment Conditioning in Diffusion-based Virtual Try-On 提出Re-CatVTON，高效单UNet扩散模型实现高性能虚拟试穿 classifier-free guidance

🔬 支柱七：动作重定向 (Motion Retargeting) (1 篇)

#题目一句话要点标签🔗⭐
37 Unsupervised Multi-View Visual Anomaly Detection via Progressive Homography-Guided Alignment 提出ViewSense-AD，通过同构变换引导对齐实现无监督多视角异常检测。 geometric consistency

🔬 支柱六：视频提取与匹配 (Video Extraction & Matching) (1 篇)

#题目一句话要点标签🔗⭐
38 MonoMSK: Monocular 3D Musculoskeletal Dynamics Estimation MonoMSK：单目视频中基于物理的3D人体骨骼肌肉动力学估计 SMPL

⬅️ 返回 cs.CV 首页 · 🏠 返回主页