Research map / Reinforcement learning / Visual and representation RL
RLSAC: Reinforcement Learning Enhanced Sample Consensus for End-to-End Robust Estimation
Chang Nie, Guangming Wang, Zhe Liu, Luca Cavalli, Marc Pollefeys, Hesheng Wang
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 |
|---|---|---|---|
| use RLSAC instead of RANSAC at high outlier rate | 2D line fitting · 2D line fitting (outlier rate 0.7) | mAA@0.5deg: 0.608 -> 0.824 | Improved |
| use RLSAC instead of RANSAC at low outlier rate | 2D line fitting · 2D line fitting (outlier rate 0.1) | mAA@0.5deg: 0.870 -> 0.875 | No clear effect |
| use RLSAC instead of MAGSAC++ | fundamental matrix estimation · Fundamental matrix estimation (1k iterations) | mAA@10deg R: 0.753 -> 0.760 | Improved |
| add 128-dim image descriptors to the state (ablation) | RLSAC · Fundamental matrix estimation | mAA@10deg R: 0.702 -> 0.760 | Improved |
| use probabilistic sampling in training and max sampling in testing (ablation) | RLSAC · Fundamental matrix estimation | mAA@10deg R: 0.730 -> 0.760 (vs max/max) | Improved |
| use probabilistic sampling in testing instead of max sampling (ablation) | RLSAC · Fundamental matrix estimation | mAA@10deg R: 0.760 -> 0.720 | Hurt |
| vary number of state points N from 150 (ablation) | RLSAC · Fundamental matrix estimation | mAA@10deg R: 0.760 (N=150) vs 0.727 (N=100), 0.747 (N=200), 0.733 (N=300) | Hurt |
| train on outlier rate 0.5 only, test on other rates (ablation) | RLSAC · 2D line fitting (outlier rate 0.6) | mAA@0.5deg: 0.845 -> 0.864 | Improved |
Closest papers by meaning
- Adversarial Diffusion for Robust Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- Distributionally Robust Reinforcement Learning with Interactive Data Collection: Fundamental Hardness and Near-Optimal Algorithms (NeurIPS 2024 · similarity 0.87)
- Blending Imitation and Reinforcement Learning for Robust Policy Improvement (ICLR 2024 · similarity 0.87)
- Robust Gymnasium: A Unified Modular Benchmark for Robust Reinforcement Learning (ICLR 2025 · similarity 0.87)
- Towards Robust Offline Reinforcement Learning under Diverse Data Corruption (ICLR 2024 · similarity 0.86)
- Sampling-Based Safe Reinforcement Learning (NeurIPS 2026 · 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.
