Research map / Detection and segmentation / Open-set and few-shot detection
Discovering Deformable Keypoint Pyramids
Jianing Qian, Anastasios Panagopoulos, Dinesh Jayaraman
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 |
|---|---|---|---|
| add the combined reconstruction loss (ablation) | Keypoint Pyramids · H3.6M keypoint regression (flattened, 30 keypoints) | RMSE: 45.23 -> 43.30 | Improved |
| add the transport loss (ablation) | Keypoint Pyramids · H3.6M keypoint regression (flattened, 30 keypoints) | RMSE: 45.85 -> 43.30 | Improved |
| add the spring loss (ablation) | Keypoint Pyramids · H3.6M keypoint regression (flattened, 30 keypoints) | RMSE: 45.28 -> 43.30 | Improved |
| use reconstruction loss overall plus level-wise transport loss instead of transport loss everywhere (ablation) | Keypoint Pyramids · H3.6M keypoint regression (flattened, 30 keypoints) | RMSE: 49.53 -> 43.30 | Improved |
| use reconstruction loss overall plus level-wise transport loss instead of reconstruction everywhere (ablation) | Keypoint Pyramids · H3.6M keypoint regression (flattened, 30 keypoints) | RMSE: 46.93 -> 43.30 | Improved |
| condition level-l keypoint encoder on level l-1 keypoints (ablation) | Keypoint Pyramids · H3.6M keypoint regression (flattened, 30 keypoints) | RMSE: 43.75 -> 43.30 | No clear effect |
| use only level 1 keypoints instead of level 2 (ablation) | Keypoint Pyramids · H3.6M keypoint regression | RMSE: 43.97 -> 52.81 | Hurt |
| flatten both levels (30 keypoints) instead of level 2 only | Keypoint Pyramids · H3.6M keypoint regression (flattened, 30 keypoints) | RMSE: 43.97 -> 43.30 | Improved |
| hierarchical Keypoint Pyramids vs flat Transporter with 30 keypoints | Transporter · H3.6M keypoint regression (flattened, 30 keypoints) | RMSE: 47.45 -> 43.30 | Improved |
| hierarchical Keypoint Pyramids vs flat KeyNet with 30 keypoints | KeyNet · H3.6M keypoint regression (flattened, 30 keypoints) | RMSE: 46.71 -> 43.30 | Improved |
| hierarchical Keypoint Pyramids (L1+L2) keypoints vs flat Transporter (30) as input to action classifier | Transporter · H3.6M action classification | accuracy: 0.177 -> 0.218 | Improved |
| hierarchical Keypoint Pyramids (L1+L2) keypoints vs flat KeyNet (30) as input to action classifier | KeyNet · H3.6M action classification | accuracy: 0.179 -> 0.218 | Improved |
Closest papers by meaning
- Unsupervised 3D Structure Inference from Category-Specific Image Collections (CVPR 2024 · similarity 0.89)
- KeyPose: Category-Level 6D Object Pose Estimation with Self-Adaptive Keypoints (AAAI 2025 · similarity 0.87)
- Meta-Point Learning and Refining for Category-Agnostic Pose Estimation (CVPR 2024 · similarity 0.87)
- Key-Grid: Unsupervised 3D Keypoints Detection using Grid Heatmap Features (NeurIPS 2024 · similarity 0.87)
- Incremental Object Keypoint Learning (CVPR 2025 · similarity 0.86)
- Back to 3D: Few-Shot 3D Keypoint Detection with Back-Projected 2D Features (CVPR 2024 · 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.
