Research map / Detection and segmentation / Open-set and few-shot detection
Self-Supervised Equivariant Learning for Oriented Keypoint Detection
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 |
|---|---|---|---|
| use rotation-equivariant layers instead of conventional convolutional layers (ablation) | oriented keypoint detector · HPatches | MMA @3px: 64.5 -> 73.1 | Helped |
| add the window-based keypoint detection loss (ablation) | oriented keypoint detector · HPatches | MMA @3px: 44.4 -> 73.1 | Helped |
| add the dense orientation alignment loss (ablation) | oriented keypoint detector · HPatches | MMA @3px: 69.8 -> 73.1 | Helped |
| use group order G36 rather than a conventional CNN (ablation) | oriented keypoint detector · HPatches | MMA @3px: 64.5 -> 73.1 | Helped |
| increase cyclic group order from G9 to G36 (ablation) | oriented keypoint detector · HPatches | MMA @3px: 62.4 -> 73.1 | Helped |
| use max pooling rather than average pooling for keypoint detection (ablation) | keypoint detection branch · HPatches | MMA @3px: 72.3 -> 73.1 | Helped |
| use learned 1x1 convolution rather than average pooling for orientation estimation (ablation) | orientation estimation branch · HPatches | MMA @3px: 70.6 -> 73.1 | Helped |
| use the proposed detector instead of SuperPoint | keypoint matching with GIFT · Extreme Rotation | PCK@5: 0.435 -> 0.476 | Helped |
| use the proposed detector instead of Key.Net | IMC2021 pose estimation, 1,024 keypoints · IMC2021 validation | mAA (5°): 0.397 -> 0.441 | Helped |
| apply orientation-based outlier filtering | ours + HardNet · HPatches | MMA @3px: 73.1 -> 76.7 | Helped |
| replace the baseline orientation with the proposed estimated orientation | ORB detector and descriptor with outlier filtering · HPatches | MMA @3px: 42.6 -> 61.7 | Helped |
| increase keypoint count from 2,048 to 8,000 | ours · IMC2021 validation | mAA (5°): 0.521 -> 0.601 | Mixed |
Closest papers by meaning
- Learning Rotation-Equivariant Features for Visual Correspondence (CVPR 2023 · similarity 0.88)
- Neural Outlier Rejection for Self-Supervised Keypoint Learning (ICLR 2020 · similarity 0.87)
- Instance-Adaptive and Geometric-Aware Keypoint Learning for Category-Level 6D Object Pose Estimation (CVPR 2024 · similarity 0.84)
- Key-Grid: Unsupervised 3D Keypoints Detection using Grid Heatmap Features (NeurIPS 2024 · similarity 0.84)
- FRED: Towards a Full Rotation-Equivariance in Aerial Image Object Detection (AAAI 2024 · similarity 0.83)
- Towards Complementary Keypoint Detection via Mixture of Detectors (NeurIPS 2026 · similarity 0.83)
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.