cs.CV（2024-12-19）

📊 共 44 篇论文 | 🔗 19 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (13 🔗6) 支柱九：具身大模型 (Embodied Foundation Models) (12 🔗4) 支柱三：空间感知与语义 (Perception & Semantics) (9 🔗5) 支柱一：机器人控制 (Robot Control) (6 🔗3) 支柱四：生成式动作 (Generative Motion) (3) 支柱六：视频提取与匹配 (Video Extraction) (1 🔗1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (13 篇)

#题目一句话要点标签🔗⭐
1 Multimodal Hypothetical Summary for Retrieval-based Multi-image Question Answering 提出多模态假设摘要(MHyS)方法，提升检索式多图问答性能。 contrastive learning large language model multimodal
2 MMAudio: Taming Multimodal Joint Training for High-Quality Video-to-Audio Synthesis MMAudio：提出多模态联合训练框架，用于高质量视频到音频合成 flow matching multimodal ✅
3 DiffSim: Taming Diffusion Models for Evaluating Visual Similarity DiffSim：利用扩散模型评估视觉相似性，提升生成模型质量 contrastive learning diff-sim
4 SqueezeMe: Mobile-Ready Distillation of Gaussian Full-Body Avatars SqueezeMe：高斯全身体化身移动端实时蒸馏框架 distillation splatting ✅
5 Flowing from Words to Pixels: A Noise-Free Framework for Cross-Modality Evolution 提出CrossFlow，一种无需噪声的跨模态Flow Matching框架，实现模态间的直接映射。 flow matching depth estimation classifier-free guidance
6 {S$^3$-Mamba}: Small-Size-Sensitive Mamba for Lesion Segmentation 提出S$^3$-Mamba，提升Mamba模型对小病灶分割的敏感性，助力早期疾病诊断。 Mamba curriculum learning ✅
7 OnlineVPO: Align Video Diffusion Model with Online Video-Centric Preference Optimization OnlineVPO：通过在线视频偏好优化对齐视频扩散模型，提升生成质量。 policy learning preference learning DPO
8 GURecon: Learning Detailed 3D Geometric Uncertainties for Neural Surface Reconstruction GURecon：学习神经表面重建的细粒度3D几何不确定性 distillation geometric consistency ✅
9 SCKD: Semi-Supervised Cross-Modality Knowledge Distillation for 4D Radar Object Detection 提出SCKD半监督跨模态知识蒸馏方法，提升4D雷达目标检测性能。 distillation ✅
10 Scaling 4D Representations 通过扩展4D表征，显著提升视频自监督学习在时空任务上的性能 MAE depth estimation ✅
11 Learning Visual Composition through Improved Semantic Guidance 通过改进语义指导提升视觉组合学习能力，显著增强CLIP模型性能。 representation learning contrastive learning
12 Prompt-A-Video: Prompt Your Video Diffusion Model via Preference-Aligned LLM Prompt-A-Video：提出基于偏好对齐LLM的视频扩散模型Prompt优化框架 DPO direct preference optimization
13 Token Preference Optimization with Self-Calibrated Visual-Anchored Rewards for Hallucination Mitigation 提出自校准视觉锚定奖励的Token偏好优化以缓解幻觉问题 DPO direct preference optimization

🔬 支柱九：具身大模型 (Embodied Foundation Models) (12 篇)

#题目一句话要点标签🔗⭐
14 Unveiling Uncertainty: A Deep Dive into Calibration and Performance of Multimodal Large Language Models 研究多模态大语言模型校准问题，提出IDK数据集并优化提示以提升不确定性评估。 large language model multimodal ✅
15 OpenEMMA: Open-Source Multimodal Model for End-to-End Autonomous Driving OpenEMMA：开源多模态大模型，用于端到端自动驾驶 large language model multimodal chain-of-thought ✅
16 WikiStyle+: A Multimodal Approach to Content-Style Representation Disentanglement for Artistic Image Stylization 提出WikiStyle+数据集和解耦扩散模型，实现多模态艺术图像风格迁移 multimodal
17 Multi-QuAD: Multi-Level Quality-Adaptive Dynamic Network for Reliable Multimodal Classification 提出Multi-QuAD，解决多模态分类中样本质量差异导致的可靠性问题 multimodal
18 MegaPairs: Massive Data Synthesis For Universal Multimodal Retrieval MegaPairs：大规模数据合成，用于通用多模态检索 multimodal
19 Multimodal Latent Diffusion Model for Complex Sewing Pattern Generation 提出 SewingLDM，用于生成受文本、体型和草图控制的复杂缝纫图案 multimodal ✅
20 FedPIA -- Permuting and Integrating Adapters leveraging Wasserstein Barycenters for Finetuning Foundation Models in Multi-Modal Federated Learning FedPIA：利用Wasserstein重心置换和集成适配器，用于多模态联邦学习中微调基础模型 foundation model
21 FiVL: A Framework for Improved Vision-Language Alignment through the Lens of Training, Evaluation and Explainability FiVL：通过训练、评估和可解释性增强视觉语言模型中的视觉对齐 multimodal visual grounding ✅
22 Movie2Story: A framework for understanding videos and telling stories in the form of novel text 提出MSBench基准测试，用于评估多模态大语言模型在长视频故事生成中的能力 large language model
23 TextSleuth: Towards Explainable Tampered Text Detection 提出TextSleuth以解决可解释的篡改文本检测问题 multimodal
24 HiCM$^2$: Hierarchical Compact Memory Modeling for Dense Video Captioning 提出HiCM²以解决密集视频字幕生成问题 large language model
25 Llama Learns to Direct: DirectorLLM for Human-Centric Video Generation 提出DirectorLLM，利用LLM编排人体姿态，提升人本视频生成质量 large language model

🔬 支柱三：空间感知与语义 (Perception & Semantics) (9 篇)

#题目一句话要点标签🔗⭐
26 LiHi-GS: LiDAR-Supervised Gaussian Splatting for Highway Driving Scene Reconstruction 提出LiHi-GS，利用激光雷达监督的高斯溅射重建高速公路驾驶场景 gaussian splatting splatting NeRF ✅
27 GSRender: Deduplicated Occupancy Prediction via Weakly Supervised 3D Gaussian Splatting GSRender：基于弱监督3D高斯溅射的去重占用预测，提升自动驾驶感知性能 3D gaussian splatting gaussian splatting splatting ✅
28 Affordance-Aware Object Insertion via Mask-Aware Dual Diffusion 提出Mask-Aware Dual Diffusion模型，用于可控的、符合常识的物体插入 affordance affordance-aware ✅
29 Improving Geometry in Sparse-View 3DGS via Reprojection-based DoF Separation 提出基于重投影的自由度分离方法，提升稀疏视角3DGS的几何重建质量 3D gaussian splatting 3DGS gaussian splatting
30 Bright-NeRF:Brightening Neural Radiance Field with Color Restoration from Low-light Raw Images Bright-NeRF：提出色彩恢复的神经辐射场，解决低光照RAW图像的新视角合成问题 NeRF neural radiance field
31 SolidGS: Consolidating Gaussian Surfel Splatting for Sparse-View Surface Reconstruction SolidGS：通过巩固高斯Surfel Splatting实现稀疏视角下的表面重建 gaussian splatting splatting
32 ObjVariantEnsemble: Advancing Point Cloud LLM Evaluation in Challenging Scenes with Subtly Distinguished Objects ObjVariantEnsemble：针对细微差异物体的点云LLM评测基准 scene understanding spatial relationship embodied AI
33 LiDAR-RT: Gaussian-based Ray Tracing for Dynamic LiDAR Re-simulation 提出LiDAR-RT，利用高斯基元光线追踪实现动态LiDAR实时重仿真 neural radiance field ✅
34 PC-BEV: An Efficient Polar-Cartesian BEV Fusion Framework for LiDAR Semantic Segmentation 提出PC-BEV，高效融合极坐标与笛卡尔坐标BEV特征，用于LiDAR语义分割 scene understanding ✅

🔬 支柱一：机器人控制 (Robot Control) (6 篇)

#题目一句话要点标签🔗⭐
35 Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representations 提出基于视频预测策略（VPP）的通用机器人策略，利用预测视觉表征提升机器人操作能力。 manipulation dexterous manipulation contrastive learning
36 Can We Get Rid of Handcrafted Feature Extractors? SparseViT: Nonsemantics-Centered, Parameter-Efficient Image Manipulation Localization through Spare-Coding Transformer SparseViT：通过稀疏编码Transformer自适应提取非语义特征，实现高效图像篡改定位 manipulation
37 Arti-PG: A Toolbox for Procedurally Synthesizing Large-Scale and Diverse Articulated Objects with Rich Annotations Arti-PG：用于程序化合成大规模、多样化、带丰富标注的铰接物体工具箱 manipulation affordance
38 A Light-Weight Framework for Open-Set Object Detection with Decoupled Feature Alignment in Joint Space 提出轻量级解耦开放集目标检测框架DOSOD，提升机器人应用中的实时性。 manipulation ✅
39 Efficient Neural Network Encoding for 3D Color Lookup Tables 提出高效神经网络编码以压缩3D颜色查找表 manipulation ✅
40 TDCNet: Transparent Objects Depth Completion with CNN-Transformer Dual-Branch Parallel Network 提出TDCNet，利用CNN-Transformer双分支并行网络完成透明物体深度补全 manipulation ✅

🔬 支柱四：生成式动作 (Generative Motion) (3 篇)

#题目一句话要点标签🔗⭐
41 EnergyMoGen: Compositional Human Motion Generation with Energy-Based Diffusion Model in Latent Space EnergyMoGen：基于能量的扩散模型在潜在空间中进行组合式人体运动生成 text-to-motion motion generation motion latent
42 ScaMo: Exploring the Scaling Law in Autoregressive Motion Generation Model ScaMo：探索自回归运动生成模型中的缩放定律 motion generation motion tokenizer
43 Jet: A Modern Transformer-Based Normalizing Flow Jet：一种基于Transformer的现代化归一化流模型，提升图像生成质量。 VQ-VAE

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

#题目一句话要点标签🔗⭐
44 GenHMR: Generative Human Mesh Recovery GenHMR：提出一种生成式人体网格恢复框架，有效应对单目图像三维重建的不确定性。 human mesh recovery HMR ✅

⬅️ 返回 cs.CV 首页 · 🏠 返回主页