Research map / 3D vision / Human pose estimation
EdgeCape: Edge Weight Prediction For Category-Agnostic Pose Estimation
Or Hirschorn, Shai Avidan
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 pose-graph prediction (predicted weighted edges) to GraphCape (ablation) | GraphCape · MP-100 split 1, 1-shot | mPCK: 79.87 -> 80.76 | Improved |
| add Markov Attention Bias to GraphCape (ablation) | GraphCape · MP-100 split 1, 1-shot | mPCK: 79.87 -> 80.78 | Improved |
| combine pose-graph prediction with Markov Attention Bias (ablation) | EdgeCape · MP-100 split 1, 1-shot | mPCK: 79.87 -> 81.96 | Improved |
| predict weighted pose-graphs instead of using fixed unweighted graph | GraphCape · MP-100 1-shot average | mPCK: 75.73 -> 77.72 | Improved |
| predict weighted pose-graphs instead of using fixed unweighted graph (PCK@0.2, saturated) | GraphCape · MP-100 5-shot average | PCK@0.2: 91.50 -> 91.87 | Improved |
| supervise adjacency with decoder-based masking instead of auxiliary GCN reconstruction (ablation) | EdgeCape · MP-100 split 1, 1-shot | PCK@0.2: 92.88 -> 93.57 | Improved |
| predict graph from keypoint features only instead of keypoint plus full image features (ablation) | pose-graph predictor · MP-100 split 1, 1-shot | PCK: -0.24% | Hurt |
| use MLP or multi-head attention edge predictors instead of cosine similarity (ablation) | pose-graph predictor · MP-100 | PCK | No clear effect |
| use DINOv2 instead of SwinV2 backbone (ablation) | EdgeCape · MP-100 split 1, 1-shot | mPCK: 80.63 -> 81.96 | Improved |
| inject up to 16 random noisy edges in input graph (ablation) | EdgeCape vs GraphCape · MP-100 noisy prior graph | PCK | Improved |
| use EdgeCape instead of GraphCape in super cross-category setting | EdgeCape · MP-100 super cross-category | PCK@0.2: Human Body 88.38 -> 91.38; Vehicle 44.06 -> 44.12; Furniture 45.56 -> 54.03 | Mixed |
| use fully-connected prior graph (no topology info) (ablation) | GraphCape · MP-100 | PCK | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SCAPE — SCAPE: A Simple and Strong Category-Agnostic Pose Estimator (ECCV 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- CapeX: Category-Agnostic Pose Estimation from Textual Point Explanation (ICLR 2025 · similarity 0.91)
- Meta-Point Learning and Refining for Category-Agnostic Pose Estimation (CVPR 2024 · similarity 0.88)
- ESCAPE: Encoding Super-keypoints for Category-Agnostic Pose Estimation (CVPR 2024 · similarity 0.87)
- Learning to Filter Outlier Edges in Global SfM (CVPR 2025 · similarity 0.85)
- SecondPose: SE(3)-Consistent Dual-Stream Feature Fusion for Category-Level Pose Estimation (CVPR 2024 · similarity 0.85)
- Recurrent Feature Mining and Keypoint Mixup Padding for Category-Agnostic Pose Estimation (CVPR 2025 · similarity 0.85)
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.
