Research map / 3D vision / Human pose estimation
Point-Set Anchors for Object Detection, Instance Segmentation and Pose Estimation
Fangyun Wei, Xiao Sun, Hongyang Li, Jingdong Wang, Stephen Lin
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 |
|---|---|---|---|
| use Corner Point with Projection matching instead of Nearest Point (ablation) | PointSetNet · COCO minival instance segmentation | mask AP: 21.9 -> 27.0 | Improved |
| use Corner Point with Projection matching instead of Nearest Line (ablation) | PointSetNet · COCO minival instance segmentation | mask AP: 23.2 -> 27.0 | Improved |
| increase anchor points from 24 to 72 (ablation) | PointSetNet · COCO minival instance segmentation | mask AP: 26.7 -> 28.0 | Improved |
| regress from point-set anchors instead of center point (ablation) | PointSetNet · COCO minival instance segmentation | mask AP: 26.0 -> 27.0 | Improved |
| use mean pose anchor instead of center point anchor (ablation) | PointSetNet · COCO val2017 keypoints | AP: 16.9 -> 40.9 | Improved |
| use mean pose anchor instead of rectangle anchor (ablation) | PointSetNet · COCO val2017 keypoints | AP: 12.0 -> 40.9 | Improved |
| use 5 K-means canonical poses instead of mean pose (ablation) | PointSetNet · COCO val2017 keypoints | AP: 40.9 -> 43.8 | Improved |
| go from 5 to 7 K-means pose anchors (ablation) | PointSetNet · COCO val2017 keypoints | AP: 43.8 -> 43.8 | No clear effect |
| widen scale range of pose anchors from 0.6:1.4 to 0.4:1.6 (ablation) | PointSetNet · COCO val2017 keypoints | AP: 42.6 -> 39.4 | Hurt |
| widen rotation range of pose anchors from -20:20 to -30:30 (ablation) | PointSetNet · COCO val2017 keypoints | AP: 43.5 -> 42.6 | Hurt |
| use deep shape indexed feature instead of box region feature (9 points) (ablation) | PointSetNet · COCO val2017 keypoints | AP: 42.8 -> 46.0 | Improved |
| apply shape indexed feature only to classification subnetwork (ablation) | PointSetNet · COCO val2017 keypoints | AP: 41.6 -> 46.0 | Mixed |
Closest papers by meaning
- RepPoints: Point Set Representation for Object Detection (ICCV 2019 · similarity 0.88)
- ObjectBox: From Centers to Boxes for Anchor-Free Object Detection (ECCV 2022 · similarity 0.87)
- Region Proposal by Guided Anchoring (CVPR 2019 · similarity 0.87)
- PointOBB: Learning Oriented Object Detection via Single Point Supervision (CVPR 2024 · similarity 0.85)
- Point2RBox: Combine Knowledge from Synthetic Visual Patterns for End-to-end Oriented Object Detection with Single Point Supervision (CVPR 2024 · similarity 0.85)
- Optimizing Human Pose Estimation Through Focused Human and Joint Regions (AAAI 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.
