Research map / Detection and segmentation / Weakly supervised detection and segmentation
Points As Queries: Weakly Semi-Supervised Object Detection by Points
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 |
|---|---|---|---|
| Encode point position and category as object queries (ablation) | Point DETR · COCO 20% data setting | AP: 31.3 -> 33.3 | Helped |
| Add positional embeddings to category-only point queries (ablation) | Point DETR · COCO 20% data setting | AP: 14.7 -> 33.3 | Helped |
| Add category embeddings to position-only point queries (ablation) | Point DETR · COCO 20% data setting | AP: 31.3 -> 33.3 | Helped |
| Use Point DETR as teacher instead of FCOS† (ablation) | FCOS student · COCO 20% data setting | AP: 31.3 -> 33.3 | Helped |
| Use Point DETR as teacher instead of FCOS† (ablation) | RetinaNet student · COCO 20% data setting | AP: 30.4 -> 32.5 | Helped |
| Use Point DETR instead of Faster R-CNN† (ablation) | Point-supervised detection · COCO 20% data setting | AP: 31.4 -> 33.3 | Helped |
| Add point-labeled COCO-80 data to training (ablation) | Point DETR · COCO-35 fully labeled; COCO-80 point labeled | AP: +3.4 points | Helped |
| Use center points instead of arbitrary points on objects (ablation) | Point DETR · COCO 20% data setting | AP: 33.3 -> 33.3 | No clear effect |
| Use point annotations to query boxes instead of DETR self-training without points (ablation) | Point DETR · COCO 20% data setting | AR100: 33.6 -> 44.5 | Helped |
| Use point annotations to query boxes instead of DETR self-training without points (ablation) | Point DETR · COCO 20% data setting | AP: 26.9 -> 26.8 | No clear effect |
| Use point annotations to query boxes instead of DETR self-training without points (ablation) | Point DETR · COCO 20% data setting | AP75: 27.5 -> 24.2 | Hurt |
| Predict box offsets relative to the point rather than directly predicting absolute boxes (ablation) | Point DETR · COCO 20% data setting | point-to-box mismatch | Helped |
Closest papers by meaning
- Relational Matching for Weakly Semi-Supervised Oriented Object Detection (CVPR 2024 · similarity 0.91)
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 2024 · similarity 0.89)
- Leveraging Imagery Data with Spatial Point Prior for Weakly Semi-supervised 3D Object Detection (AAAI 2024 · similarity 0.89)
- PointOBB-v2: Towards Simpler, Faster, and Stronger Single Point Supervised Oriented Object Detection (ICLR 2025 · similarity 0.87)
- Point2RBox-v2: Rethinking Point-supervised Oriented Object Detection with Spatial Layout Among Instances (CVPR 2025 · similarity 0.87)
- Weakly Supervised Few-Shot Object Detection with DETR (AAAI 2024 · similarity 0.87)
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.