三维视觉论文趋势 2023–2026
三维重建与神经渲染
统计范围：NeurIPS、ICLR、ICML、CVPR 四个会议，2023–2026 年，共 3,226 篇被归入该主题的论文。
主要结论
- 三维视觉论文从 2023 年的 563 篇增长到 2026 年的 1,010 篇，三年间扩大到约 1.8 倍。
- CVPR 是发三维视觉论文最多的会议，2023–2026 年共 1,641 篇，占四个会议总量的 51%。
- 交叉方向里涨得最快的是高斯过程：同时属于两个主题的论文从 2023 年的 0 篇增至 2026 年的 158 篇。
- 在「计算机视觉」的 5 个主题里，它的 2026 年论文数排第 2；同组增长最快的视频理解是 2.6 倍，三维视觉是 1.8 倍。
各会议论文数
按年份和会议统计
| 年份 | NeurIPS | ICLR | ICML | CVPR | 合计 |
|---|---|---|---|---|---|
| 2023 | 152 | 56 | 22 | 333 | 563 |
| 2024 | 222 | 96 | 44 | 389 | 751 |
| 2025 | 254 | 162 | 52 | 434 | 902 |
| 2026 | 211 | 197 | 117 | 485 | 1,010 |
代表论文
按年份和会议均匀挑选，方便快速看到每一届的样貌
- NeurIPS 2026$\alpha$Depth: Learning Single-Pass Soft Boundary Decomposition for Stereo Conversion
- ICML 20263D MeanFlow: One-Step Point Cloud Completion and Generation via Average-Velocity Transport
- ICLR 2026$\pi^3$: Permutation-Equivariant Visual Geometry Learning
- CVPR 20262D-LFM Lifting Foundation Model without 3D Supervision
- NeurIPS 20251000+ FPS 4D Gaussian Splatting for Dynamic Scene Rendering
- ICML 20253D-LMVIC: Learning-based Multi-View Image Compression with 3D Gaussian Geometric Priors
- ICLR 20253D StreetUnveiler with Semantic-aware 2DGS - a simple baseline
- CVPR 20253D Convex Splatting: Radiance Field Rendering with 3D Smooth Convexes
- NeurIPS 20243D Gaussian Rendering Can Be Sparser: Efficient Rendering via Learned Fragment Pruning
- ICML 20243D Geometric Shape Assembly via Efficient Point Cloud Matching
- ICLR 20243D Feature Prediction for Masked-AutoEncoder-Based Point Cloud Pretraining
- CVPR 2024DaReNeRF: Direction-aware Representation for Dynamic Scenes
- NeurIPS 2023[Re] Masked Autoencoders Are Small Scale Vision Learners: A Reproduction Under Resource Constraints
- ICML 2023Efficient Graph Field Integrators Meet Point Clouds
- ICLR 2023$\mathrm{SE}(3)$-Equivariant Attention Networks for Shape Reconstruction in Function Space
- CVPR 20233D Cinemagraphy From a Single Image
- NeurIPS 2026$R^3$: 3D Reconstruction via Relative Regression
- ICML 20263DGS-HPC: Distractor-free 3D Gaussian Splatting with Hybrid Patch-wise Classification
- ICLR 20263DGEER: 3D Gaussian Rendering Made Exact and Efficient for Generic Cameras
- CVPR 20263D Gaussian Splatting at Arbitrary Resolutions with Compact Proxy Anchors
全部三维视觉论文
共 197 篇，第 4 / 4 页
- ICLR 2026SpaceControl: Introducing Test-Time Spatial Control to 3D Generative Modeling
- ICLR 2026Sparkle: A Robust and Versatile Representation for Point Cloud-based Human Motion Capture
- ICLR 2026SpikeGen: Decoupled “Rods and Cones” Visual Representation Processing with Latent Generative Framework
- ICLR 2026Spiking Discrepancy Transformer for Point Cloud Analysis
- ICLR 2026Splat and Distill: Augmenting Teachers with Feed-Forward 3D Reconstruction For 3D-Aware Distillation
- ICLR 2026Splat Regression Models
- ICLR 2026Splat the Net: Radiance Fields with Splattable Neural Primitives
- ICLR 2026SSD-GS: Scattering and Shadow Decomposition for Relightable 3D Gaussian Splatting
- ICLR 2026station2radar: query‑conditioned gaussian splatting for precipitation field
- ICLR 2026STream3R: Scalable Sequential 3D Reconstruction with Causal Transformer
- ICLR 2026Streaming Visual Geometry Transformer
- ICLR 2026StreamSplat: Towards Online Dynamic 3D Reconstruction from Uncalibrated Video Streams
- ICLR 2026Stylos: Multi-View 3D Stylization with Single-Forward Gaussian Splatting
- ICLR 2026SurfSplat: Conquering Feedforward 2D Gaussian Splatting with Surface Continuity Priors
- ICLR 2026Test-Time Optimization of 3D Point Cloud LLM via Manifold-Aware In-Context Guidance and Refinement
- ICLR 2026Text-to-3D by Stitching a Multi-view Reconstruction Network to a Video Generator
- ICLR 2026The Less You Depend, The More You Learn: Synthesizing Novel Views from Sparse, Unposed Images without Any 3D Knowledge
- ICLR 2026TIGaussian: Disentangle Gaussians for Spatial-Awared Text-Image-3D Alignment
- ICLR 2026TINKER: Diffusion's Gift to 3D--Multi-View Consistent Editing From Sparse Inputs without Per-Scene Optimization
- ICLR 2026To Augment or Not to Augment? Diagnosing Distributional Symmetry Breaking
- ICLR 2026To View Transform or Not to View Transform: NeRF-based Pre-training Perspective
- ICLR 2026Towards Physically Executable 3D Gaussian for Embodied Navigation
- ICLR 2026True Self-Supervised Novel View Synthesis is Transferable
- ICLR 2026TTT3R: 3D Reconstruction as Test-Time Training
- ICLR 2026UFO-4D: Unposed Feedforward 4D reconstruction from Two Images
- ICLR 2026ULTRA-360: Unconstrained Dataset for Large-scale Temporal 3D Reconstruction across Altitudes and Omnidirectional Views
- ICLR 2026UltraGauss: Ultrafast Gaussian Reconstruction of 3D Ultrasound Volumes
- ICLR 2026Uncertainty Matters in Dynamic Gaussian Splatting for Monocular 4D Reconstruction
- ICLR 2026Uncertainty-Aware 3D Reconstruction for Dynamic Underwater Scenes
- ICLR 2026Uncertainty-Aware Gaussian Map for Vision-Language Navigation
- ICLR 2026Unified 3D Scene Understanding Through Physical World Modeling
- ICLR 2026UniSplat: Unified Spatio-Temporal Fusion via 3D Latent Scaffolds for Dynamic Driving Scene Reconstruction
- ICLR 2026UniUGG: Unified 3D Understanding and Generation via Geometric-Semantic Encoding
- ICLR 2026Universal Beta Splatting
- ICLR 2026UP2You: Fast Reconstruction of Yourself from Unconstrained Photo Collections
- ICLR 2026UrbanGS: Efficient and Scalable Architecture for Geometrically Accurate Large-Scene Reconstruction
- ICLR 2026Variation-aware Flexible 3D Gaussian Editing
- ICLR 2026VaseVQA-3D: Benchmarking 3D VLMs on Ancient Greek Pottery
- ICLR 2026VER: Vision Expert Transformer for Robot Learning via Foundation Distillation and Dynamic Routing
- ICLR 2026Vid-LLM: A Compact Video-based 3D Multimodal LLM with Reconstruction–Reasoning Synergy
- ICLR 2026VisionLaw: Inferring Interpretable Intrinsic Dynamics from Visual Observations via Bilevel Optimization
- ICLR 2026VoMP: Predicting Volumetric Mechanical Property Fields
- ICLR 2026Vulcan: Crafting Compact Class-Specific Vision Transformers For Edge Intelligence
- ICLR 2026Weight Space Representation Learning on Diverse NeRF Architectures
- ICLR 2026WinT3R: Window-Based Streaming Reconstruction with Camera Token Pool
- ICLR 2026WorldTree: Towards 4D Dynamic Worlds from Monocular Video using Tree-Chains
- ICLR 2026YoNoSplat: You Only Need One Model for Feedforward 3D Gaussian Splatting
在 Claude、Cursor 里直接问这个趋势
接上 Paperlist MCP 后，写综述、挑方向或准备投稿时，不用离开对话框就能拿到这页的数字：每年各会议的论文数、增长倍数、涨得最快的子方向和代表论文，AI 会按你的问题整理。
可以直接复制这句话问 AI
用 paperlist 查三维视觉在 CVPR 2023–2026 的论文趋势，哪一年涨得最快？
AI 会调用的工具：get_topic_stats {"topic": "3d-vision"}
常见问题
- 2023–2026 年四大会议一共有多少篇三维视觉论文？
- 共 3,226 篇，按年份分别是 2023 年 563 篇、2024 年 751 篇、2025 年 902 篇、2026 年 1,010 篇。
- NeurIPS 2026 有多少篇三维视觉论文？
- NeurIPS 2026 有 211 篇三维视觉论文，上一届是 254 篇。
- ICLR 2026 有多少篇三维视觉论文？
- ICLR 2026 有 197 篇三维视觉论文，上一届是 162 篇。
- ICML 2026 有多少篇三维视觉论文？
- ICML 2026 有 117 篇三维视觉论文，上一届是 52 篇。
- CVPR 2026 有多少篇三维视觉论文？
- CVPR 2026 有 485 篇三维视觉论文，上一届是 434 篇。
- 哪个会议的三维视觉论文最多？
- CVPR，2023–2026 年共 1,641 篇。