Research map / 3D vision / Depth and stereo
Learning Efficient Photometric Feature Transform for Multi-View Stereo
Kaizhang Kang, Cihui Xie, Ruisheng Zhu, Xiaohe Ma, Ping Tan, Hongzhi Wu, Kun 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 intensity-insensitive branch to intensity-sensitive branch (bandwidth 6) (ablation) | photometric feature network · training loss | L_main: 53.98 -> 11.33 | Improved |
| add intensity-insensitive branch (bandwidth 8) (ablation) | photometric feature network · training loss | L_main: 53.31 -> 10.33 | Improved |
| add intensity-insensitive branch (bandwidth 10) (ablation) | photometric feature network · training loss | L_main: 53.17 -> 9.45 | Improved |
| increase number of measurements (bandwidth 6 to 10) (ablation) | full network · training loss | L_main: 11.33 -> 9.45 | Improved |
| use learned photometric features instead of CasMVSNet | multi-view stereo on lightstage data · Fox | completeness at 1mm (%): 34.2 -> 71.9 | Improved |
| use learned features instead of COLMAP on full-on lighting images | COLMAP pipeline · Cat | completeness at 1mm (%): 05.7 -> 83.6 | Improved |
| use learned features instead of COLMAP on full-on lighting images | COLMAP pipeline · Rooster | accuracy at 1mm (%): 70.6 -> 65.3 | Hurt |
| use learned features instead of COLMAP on full-on lighting images | COLMAP pipeline · Cup | accuracy at 1mm (%): 51.4 -> 44.6 | Hurt |
| add intensity-insensitive branch to sensitive-only network (ablation) | photometric feature network · synthetic ball | feature discrimination | Improved |
| add noise with 5 times training magnitude (5% vs 1%) | learned features · geometric reconstruction | reconstruction quality | No clear effect |
Closest papers by meaning
- MVPSNet: Fast Generalizable Multi-view Photometric Stereo (ICCV 2023 · similarity 0.89)
- Neural Multi-View Self-Calibrated Photometric Stereo without Photometric Stereo Cues (ICCV 2025 · similarity 0.89)
- Learning a Multi-View Stereo Machine (NeurIPS 2017 · similarity 0.88)
- Sparse Views Near Light: A Practical Paradigm for Uncalibrated Point-light Photometric Stereo (CVPR 2024 · similarity 0.88)
- MVCPS-NeuS: Multi-view Constrained Photometric Stereo for Neural Surface Reconstruction (CVPR 2024 · similarity 0.86)
- Differentiable Display Photometric Stereo (CVPR 2024 · 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.
