Research map / 3D vision / Depth and stereo
Take Your Model Further: A General Post-refinement Network for Light Field Disparity Estimation via BadPix Correction
Rongshan Chen, Hao Sheng, Da Yang, Sizhe Wang, Zhenglong Cui, Ruixuan Cong
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 |
|---|---|---|---|
| Apply BpCNet refinement to EPI-Shift | EPI-Shift · HCI 4D Light Field Dataset, average | BadPix(0.03): 32.19-21.00-2.73 | Improved |
| Apply BpCNet refinement to LFattNet | LFattNet · HCI 4D Light Field Dataset, average | BadPix(0.01): 15.07-1.34 | Improved |
| Apply BpCNet refinement to DistgDisp | DistgDisp · HCI 4D Light Field Dataset, average | BadPix(0.01): 21.71+2.10-2.29 | Mixed |
| Apply BpCNet refinement to LFattNet | LFattNet · HCI 4D Light Field Dataset, Dots | BadPix(0.01): +2.86 | Hurt |
| Add LF image augmentation (ablation) | BpCNet + SPO · HCI 4D Light Field Dataset, average | BadPix(0.07): -3.42 -> -3.85 | Improved |
| Add initial disparity augmentation (ablation) | BpCNet + SPO · HCI 4D Light Field Dataset, average | BadPix(0.07): -2.33 -> -3.85 | Improved |
| Add iterative training (ablation) | BpCNet + SPO · HCI 4D Light Field Dataset, average | BadPix(0.07): -3.74 -> -3.85 | Improved |
| Increase hypothesis disparity count from K=3 to K=9 (ablation) | BpCNet + SPO · HCI 4D Light Field Dataset, average | BadPix(0.07): -1.03 -> -3.85 | Improved |
| Use phase-shift warp instead of bilinear warp (ablation) | BpCNet + SPO · HCI 4D Light Field Dataset, average | BadPix(0.07): -3.67 -> -3.85 | Improved |
| Use phase-shift warp instead of bicubic warp (ablation) | BpCNet + SPO · HCI 4D Light Field Dataset, average | BadPix(0.07): -3.74 -> -3.85 | Improved |
| Use the full BpCNet rather than the simple refinement model (ablation) | BpCNet + SPO · HCI 4D Light Field Dataset, average | BadPix(0.07): +5.32 -> -3.85 | Improved |
| Apply an EPINet-trained BpCNet to other algorithms without specific training | Other disparity algorithms · HCI 4D Light Field Dataset | BadPix(0.07): 2.98-0.24 | Improved |
Closest papers by meaning
- Fast Light-Field Disparity Estimation With Multi-Disparity-Scale Cost Aggregation (ICCV 2021 · similarity 0.89)
- Learning Fused Pixel and Feature-Based View Reconstructions for Light Fields (CVPR 2020 · similarity 0.87)
- OPAL: Occlusion Pattern Aware Loss for Unsupervised Light Field Disparity Estimation (TPAMI 2023 · similarity 0.87)
- What is Learned in Deep Uncalibrated Photometric Stereo? (ECCV 2020 · similarity 0.87)
- PX-NET: Simple and Efficient Pixel-Wise Training of Photometric Stereo Networks (ICCV 2021 · similarity 0.87)
- Disentangling Light Fields for Super-Resolution and Disparity Estimation (TPAMI 2022 · similarity 0.87)
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.
