Research map / 3D vision / Depth and stereo
Deep Material-Aware Cross-Spectral Stereo Matching
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 simultaneous disparity prediction and RGB-to-NIR translation | Cross-spectral stereo matching · RGB-NIR stereo dataset, Mean | disparity RMSE (pixels): 1.28 -> 0.80 | Helped |
| Use RGB-NIR stereo input rather than only RGB input (ablation) | DPN · RGB-NIR stereo dataset, Mean | disparity RMSE (pixels): 1.06 -> 0.80 | Helped |
| Learn the RGB-to-NIR translation rather than averaging RGB channels (ablation) | STN · RGB-NIR stereo dataset, Mean | disparity RMSE (pixels): 0.89 -> 0.80 | Helped |
| Use a symmetric CNN in the STN to prevent learning disparity (ablation) | STN · RGB-NIR stereo dataset, Mean | disparity RMSE (pixels): 0.89 -> 0.80 | Helped |
| Add material-aware losses (ablation) | DPN · RGB-NIR stereo dataset, Mean | disparity RMSE (pixels): 1.00 -> 0.80 | Helped |
| Add light-specific loss handling (ablation) | DPN · Light | disparity RMSE (pixels): 0.81 -> 0.69 | Helped |
| Add glass-specific loss handling (ablation) | DPN · Glass | disparity RMSE (pixels): 0.97 -> 0.65 | Helped |
| Add glossy-surface-specific loss handling (ablation) | DPN · Glossy | disparity RMSE (pixels): 1.23 -> 0.70 | Helped |
| Add confidence weighting to disparity smoothing (ablation) | DPN · RGB-NIR stereo dataset, Mean | disparity RMSE (pixels): 0.87 -> 0.80 | Helped |
Closest papers by meaning
- Unsupervised Cross-Spectral Stereo Matching by Learning to Synthesize (AAAI 2019 · similarity 0.90)
- Adaptive Multi-Modal Cross-Entropy Loss for Stereo Matching (CVPR 2024 · similarity 0.86)
- Cross-Spectral Gaussian Splatting with Spatial Occupancy Consistency (AAAI 2025 · similarity 0.85)
- Learned Binocular-Encoding Optics for RGBD Imaging Using Joint Stereo and Focus Cues (CVPR 2025 · similarity 0.84)
- Cross-spectral Gated-RGB Stereo Depth Estimation (CVPR 2024 · similarity 0.84)
- Stereo Anywhere: Robust Zero-Shot Deep Stereo Matching Even Where Either Stereo or Mono Fail (CVPR 2025 · similarity 0.84)
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.