Research map / 3D vision / Gaussian splatting
PointSplat: Compact Gaussian Splatting via Human-Centric Prediction
Yujie Guo, Yudong Jin, Lingteng Qiu, Zehong Shen, Zhen Xu, Zhang Jing, Xianchao Shen, Hujun Bao, Sida Peng, Xiaowei Zhou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add voxel point encoding (ablation) | PointSplat · DNA-Rendering | PSNR: 26.60 -> 27.18 | Improved |
| add voxel point encoding (ablation) | PointSplat · DNA-Rendering | GS-Num: 160k -> 71k | Improved |
| add ray casting (ablation) | PointSplat · DNA-Rendering | PSNR: 26.32 -> 27.18 | Improved |
| add ray casting (ablation) | PointSplat · DNA-Rendering | GS-Num: 224k -> 71k | Improved |
| use alternating attention instead of full self-attention (ablation) | PointSplat · DNA-Rendering | PSNR and time: 27.09 -> 27.18; time 1.1 -> 0.4 | Improved |
| use visual hull instead of VGGT depth proxy (ablation) | PointSplat · DNA-Rendering | PSNR: 25.95 -> 27.18 | Improved |
| use visual hull instead of bounding box proxy (ablation) | PointSplat · DNA-Rendering | PSNR: 23.89 -> 27.18 | Improved |
| use visual hull instead of frustum sample proxy (ablation) | PointSplat · DNA-Rendering | PSNR: 22.92 -> 27.18 | Improved |
| reduce voxel size from 0.0050 to 0.0030 (ablation) | PointSplat · DNA-Rendering | PSNR and GS-Num: 27.18 -> 27.52; 71k -> 189k | Mixed |
| increase voxel size from 0.0050 to 0.0100 (ablation) | PointSplat · DNA-Rendering | PSNR: 27.18 -> 25.51 | Hurt |
| drop 10% of foreground mask pixels at inference (ablation) | PointSplat · DNA-Rendering | PSNR: 26.74 -> 24.55 | Hurt |
| use human-centric prediction instead of view-centric DepthSplat | PointSplat · DNA-Rendering | PSNR: 23.98 -> 27.18 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- GPS-Gaussian — GPS-Gaussian: Generalizable Pixel-wise 3D Gaussian Splatting for Real-time Human Novel View Synthesis (CVPR 2024) · better on 4 of 4 shared results
- DepthSplat — DepthSplat: Connecting Gaussian Splatting and Depth (CVPR 2025) · better on 3 of 3 shared results
- GS-LRM — GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting (ECCV 2024) · better on 3 of 3 shared results
- RoGSplat — RoGSplat: Learning Robust Generalizable Human Gaussian Splatting from Sparse Multi-View Images (CVPR 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Motion Matters: Compact Gaussian Streaming for Free-Viewpoint Video Reconstruction (NeurIPS 2025 · similarity 0.88)
- SplatFormer: Point Transformer for Robust 3D Gaussian Splatting (ICLR 2025 · similarity 0.88)
- HumanSplat: Generalizable Single-Image Human Gaussian Splatting with Structure Priors (NeurIPS 2024 · similarity 0.88)
- Guess The Unseen: Dynamic 3D Scene Reconstruction from Partial 2D Glimpses (CVPR 2024 · similarity 0.88)
- pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction (CVPR 2024 · similarity 0.88)
- GauHuman: Articulated Gaussian Splatting from Monocular Human Videos (CVPR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
