Research map / Label-efficient and robust learning / Recognition and long-tail
Unsupervised Deep Asymmetric Stereo Matching With Spatially-Adaptive Self-Similarity
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 |
|---|---|---|---|
| Adding feature-metric loss on SASS features (vs. no feature-metric loss) (ablation) | unsupervised PSMNet-hourglass stereo · KITTI 2015 resolution asymmetry | EPE: 2.652 -> 2.221 (SASS) / 2.183 (SASS+Lcs) | Helped |
| Replacing fixed FCSS sampling patterns with adaptive SASS patterns (ablation) | feature-metric loss features · KITTI 2015 resolution asymmetry | EPE: 2.329 -> 2.221 | Helped |
| Replacing fixed FCSS sampling patterns with adaptive SASS patterns (ablation) | feature-metric loss features · KITTI 2015 noise asymmetry | EPE: 2.919 -> 2.727 | Helped |
| Adding contrastive similarity loss with positive/negative weights (Lcs) on top of SASS (ablation) | SASS · KITTI 2015 resolution and noise asymmetry | EPE: resolution 2.190 -> 2.183; noise 2.683 -> 2.667 | Helped |
| Adding raw encoder feature-metric loss (no self-similarity) (ablation) | unsupervised stereo baseline · KITTI 2015 noise asymmetry | EPE: 3.614 -> 3.128 | Helped |
| Adding SASS + Lcs feature-metric loss over DAUS feature-consistency | unsupervised asymmetric stereo · KITTI 2015 resolution and noise asymmetry | EPE: resolution 2.423 -> 2.183; noise 3.071 -> 2.667 | Helped |
| Increasing number of sampling patterns L from 4 to 64 for FCSS (ablation) | FCSS self-similarity · KITTI 2015 resolution asymmetry | EPE: 2.52 -> 2.24 | Helped |
| Increasing number of sampling patterns L from 16 to 64 for SASS + Lcs (ablation) | SASS + Lcs · KITTI 2015 resolution asymmetry | EPE: 2.18 -> 2.17 | No clear effect |
| Using SASS (adaptive) instead of FCSS with the same number of patterns (ablation) | SASS features · symmetric KITTI 2015 | EPE: 1.882 (FCSS) vs 1.884 (SASS) | No clear effect |
Closest papers by meaning
- DispViT: Direct Stereo Disparity Regression with a Single-Stream Vision Transformer (ICLR 2026 · similarity 0.89)
- Any-Stereo: Arbitrary Scale Disparity Estimation for Iterative Stereo Matching (AAAI 2024 · similarity 0.89)
- Stereo Anywhere: Robust Zero-Shot Deep Stereo Matching Even Where Either Stereo or Mono Fail (CVPR 2025 · similarity 0.87)
- Consistency-aware Self-Training for Iterative-based Stereo Matching (CVPR 2025 · similarity 0.87)
- Selective-Stereo: Adaptive Frequency Information Selection for Stereo Matching (CVPR 2024 · similarity 0.87)
- Stereo Risk: A Continuous Modeling Approach to Stereo Matching (ICML 2024 · 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.