Research map / 3D vision / Human pose estimation
Learning To Find Good Models in RANSAC
Check 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 |
|---|---|---|---|
| Replace traditional model scoring with MQ-Net trained jointly on essential and fundamental matrix estimation | RANSAC relative-pose estimation · PhotoTourism, fundamental matrix | median rotation error (°): 2.33 -> 1.67 | Helped |
| Replace traditional model scoring with MQ-Net trained jointly on essential and fundamental matrix estimation | RANSAC relative-pose estimation · PhotoTourism, essential matrix | median translation error (°): 5.03 -> 2.40 | Helped |
| Add MF-Net minimal-sample filtering to MQ-Net | RANSAC relative-pose estimation · PhotoTourism, fundamental matrix | time (ms): 3.75 -> 1.65 | Helped |
| Add MF-Net minimal-sample filtering to MQ-Net | RANSAC relative-pose estimation · PhotoTourism, essential matrix | median rotation error (°): 0.94 -> 0.91 | Helped |
| Train MQ-Net jointly on essential and fundamental matrix estimation rather than only essential matrix estimation | MQ-Net · PhotoTourism, fundamental matrix | rotation mAA@10°: 0.62 -> 0.70 | Helped |
| Train MQ-Net jointly on essential and fundamental matrix estimation rather than only essential matrix estimation | MQ-Net · PhotoTourism, essential matrix | rotation mAA@10°: 0.76 -> 0.78 | Helped |
| Add an adaptive final model polishing step (ablation) | MQ-Net · PhotoTourism, essential matrix | rotation mAA@10°: 0.66 -> 0.78 | Helped |
| Use all points below ϵmax for final re-fitting rather than adaptive inlier selection (ablation) | MQ-Net · PhotoTourism, essential matrix | rotation mAA@10°: 0.78 -> 0.16 | Hurt |
| Group residuals into 10 steps during adaptive final re-fitting (ablation) | MQ-Net · PhotoTourism, essential matrix | time (ms): 27.97 -> 5.38 | Helped |
| Group residuals into 10 steps during adaptive final re-fitting (ablation) | MQ-Net · PhotoTourism, fundamental matrix | time (ms): 28.49 -> 8.42 | Helped |
| Increase histogram size from 100 to 2500 bins (ablation) | MQ-Net (E & F) · PhotoTourism, essential matrix | rotation mAA@10°: 0.77 -> 0.78 | Helped |
| Increase histogram size from 100 to 2500 bins (ablation) | MQ-Net (E & F) · PhotoTourism, fundamental matrix | rotation mAA@10°: 0.70 -> 0.64 | Hurt |
Closest papers by meaning
- RANSAC Scoring Done Right (NeurIPS 2026 · similarity 0.88)
- PARSAC: Accelerating Robust Multi-Model Fitting with Parallel Sample Consensus (AAAI 2024 · similarity 0.86)
- LAION-C: An Out-of-Distribution Benchmark for Web-Scale Vision Models (ICML 2025 · similarity 0.83)
- Learning to Filter Outlier Edges in Global SfM (CVPR 2025 · similarity 0.82)
- Scaling and Masking: A New Paradigm of Data Sampling for Image and Video Quality Assessment (AAAI 2024 · similarity 0.82)
- MRC-Net: 6-DoF Pose Estimation with MultiScale Residual Correlation (CVPR 2024 · similarity 0.82)
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.