Research map / 3D vision / Human pose estimation
A2J: Anchor-to-Joint Regression Network for 3D Articulated Pose Estimation From a Single Depth Image
Fu Xiong, Boshen Zhang, Yang Xiao, Zhiguo Cao, Taidong Yu, Joey Tianyi Zhou, Junsong Yuan
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 anchor proposal branch (ablation) | A2J · NYU Hand | mean 3D error (mm): 10.08 -> 8.61 | Improved |
| add anchor proposal branch (ablation) | A2J · ITOP front-view | mAP: 80.1 -> 88.0 | Improved |
| add informative anchor point surrounding loss (ablation) | A2J · NYU Hand | mean 3D error (mm): 9.00 -> 8.61 | Improved |
| add informative anchor point surrounding loss (ablation) | A2J · ITOP front-view | mAP: 86.4 -> 88.0 | Improved |
| use separate branches for in-plane offset and depth instead of one branch (ablation) | A2J · NYU Hand | mean 3D error (mm): 8.95 -> 8.61 | Improved |
| use separate branches for in-plane offset and depth instead of one branch (ablation) | A2J · ITOP front-view | mAP: 87.4 -> 88.0 | Improved |
| anchor-based paradigm instead of global regression (ablation) | ResNet-50 pose network · K2HPD | PDJ (0.05): 61.5 -> 76.3 | Improved |
| anchor-based paradigm instead of FCN model (ablation) | ResNet-50 pose network · K2HPD | PDJ (0.05): 70.4 -> 76.3 | Improved |
| ImageNet pre-training (ablation) | A2J · NYU Hand | mean 3D error (mm): 10.08 -> 8.61 | Improved |
| ImageNet pre-training (ablation) | A2J · ITOP front-view | mAP: 87.3 -> 88.0 | Improved |
| deeper backbone ResNet-18 to ResNet-50 (ablation) | A2J · NYU Hand | mean 3D error (mm): 9.32 -> 8.61 | Improved |
| deeper backbone ResNet-18 to ResNet-50 (ablation) | A2J · NYU Hand | FPS: 192.25 -> 105.06 | 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.
- HandPointNet — Hand PointNet: 3D Hand Pose Estimation Using Point Sets (CVPR 2018) · 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
- A2J-Transformer: Anchor-to-Joint Transformer Network for 3D Interacting Hand Pose Estimation From a Single RGB Image (CVPR 2023 · similarity 0.90)
- 3D Hand Pose Estimation via Articulated Anchor-to-Joint 3D Local Regressors (TPAMI 2025 · similarity 0.89)
- Point-to-Point Regression PointNet for 3D Hand Pose Estimation (ECCV 2018 · similarity 0.87)
- JGR-P2O: Joint Graph Reasoning based Pixel-to-Offset Prediction Network for 3D Hand Pose Estimation from a Single Depth Image (ECCV 2020 · similarity 0.87)
- V2V-PoseNet: Voxel-to-Voxel Prediction Network for Accurate 3D Hand and Human Pose Estimation From a Single Depth Map (CVPR 2018 · similarity 0.87)
- PandaPose: 3D Human Pose Lifting from a Single Image via Propagating 2D Pose Prior to 3D Anchor Space (NeurIPS 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.
