Research map / Point clouds / Point cloud classification
RPBG: Towards Robust Neural Point-based Graphics in the Wild
Qingtian Zhu, Zizhuang Wei, Zhongtian Zheng, Yifan Zhan, Zhuyu Yao, Jiawang Zhang, Kejian Wu, Yinqiang Zheng
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 Downgrade-aware Convolution (DAC) to the renderer (ablation) | RPBG renderer · T&T | PSNR: 21.88 -> 22.92 | Improved |
| add FFT loss (ablation) | RPBG renderer · T&T | SSIM: 0.769 -> 0.794 SSIM; PSNR 22.92 -> 22.94 | Improved |
| replace NPBG U-Net with restoration-style multi-scale fusion renderer (ablation) | NPBG · T&T | PSNR: 20.01 -> 21.88 | Improved |
| use multi-stage renderer instead of multi-scale fusion (ablation) | RPBG renderer · T&T | memory (GB): 12.67 -> 17.14 GB with PSNR 21.88 -> 21.75 | Hurt |
| use a learnable default feature vector for background instead of zeros (ablation) | RPBG · Caterpillar (T&T) | PSNR: 19.89 -> 21.78 | Improved |
| replace learnable vector with a 10^6-point sphere environment (ablation) | RPBG · Caterpillar (T&T) | PSNR: 21.78 -> 21.68 | No clear effect |
| add point cloud augmentation to SfM-initialized points (ablation) | RPBG · Church (SfM init) | PSNR: 21.41 -> 21.86 | Improved |
| add second round of point cloud augmentation (ablation) | RPBG · Church (SfM init) | PSNR: 21.86 -> 21.85 | No clear effect |
| use point-based re-rendering instead of RF-based methods on sparse-view scenes | RPBG vs Gaussian Splatting · GigaMVS | PSNR: 16.84 -> 20.54 | Improved |
| degrade triangulation from ground truth to random points (ablation) | RPBG · triangulation configurations | PSNR: total drop of 3.29 dB | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- NPBG — NPBG++: Accelerating Neural Point-Based Graphics (CVPR 2022) · better on 2 of 2 shared results
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
- NPBG++: Accelerating Neural Point-Based Graphics (CVPR 2022 · similarity 0.90)
- HashPoint: Accelerated Point Searching and Sampling for Neural Rendering (CVPR 2024 · similarity 0.87)
- Learning Robust Generalizable Radiance Field with Visibility and Feature Augmented Point Representation (ICLR 2024 · similarity 0.87)
- View Synthesis with Sculpted Neural Points (ICLR 2023 · similarity 0.87)
- NexusGS: Sparse View Synthesis with Epipolar Depth Priors in 3D Gaussian Splatting (CVPR 2025 · similarity 0.86)
- GoLF-NRT: Integrating Global Context and Local Geometry for Few-Shot View Synthesis (CVPR 2025 · similarity 0.86)
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.
