Research map / Detection and segmentation / Weakly supervised detection and segmentation
Omni-DETR: Omni-Supervised Object Detection With Transformers
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 |
|---|---|---|---|
| Add weak annotations to the 10% labeled-data baseline | Omni-DETR · COCO-standard 10% | mAP: 28.0 -> 34.7 (TagsU); 28.0 -> 36.8 (BoxesU) | Helped |
| Add image-level tags instead of using only unlabeled data | Omni-DETR · COCO-standard 10% | mAP: 32.4 -> 34.7 | Helped |
| Add counts to image-level tags | Omni-DETR · COCO-standard 10% | mAP: 34.7 -> 35.2; gain of 0.5% | Helped |
| Add class tags to points | Omni-DETR · COCO-standard 10% | mAP: 34.1 -> 35.7; 1.6% improvement | Helped |
| Use unified Hungarian matching instead of simple heuristic filtering (ablation) | Omni-DETR · COCO-standard 10% with weak labels | mAP: 33.3 -> 34.7 (TagsU); 33.8 -> 35.2 (TagsK); 32.4 -> 34.1 (PointsU); 34.6 -> 35.7 (PointsK) | Helped |
| Set confidence threshold τ to 0.7 rather than 0.5 (ablation) | Omni-DETR without annotations · COCO-standard 10% | mAP: 28.9 -> 32.4 | Helped |
| Set matching trade-off γ to 0.5 rather than 0.00 (ablation) | Omni-DETR with PointsK · COCO-standard 10% | mAP: 34.1 -> 35.7 | Helped |
| Use pseudo bounding boxes for unlabeled-data learning (ablation) | Omni-DETR · Experiments reported in paper | mAP: 0.5-1% | Helped |
| Use Extreme Clicking boxes rather than high-quality boxes | Omni-DETR · COCO-standard 10% | mAP: 36.4 vs. 36.8; 0.3% worse | Hurt |
| Add weak annotations to the baseline using only labeled data | Omni-DETR · COCO-standard 10% | mAP: 28.0 -> 32.4 (90% None) | Helped |
| Use Omni-DETR rather than the strongest listed baseline | COCO-standard 5% · COCO-standard 5% | mAP: 28.3 -> 30.2 | Helped |
| Use Omni-DETR rather than the strongest listed baseline | COCO-standard 10% · COCO-standard 10% | mAP: 31.6 -> 34.1 | Helped |
Closest papers by meaning
- Weakly Supervised Few-Shot Object Detection with DETR (AAAI 2024 · similarity 0.87)
- Open-Det: An Efficient Learning Framework for Open-Ended Detection (ICML 2025 · similarity 0.87)
- MS-DETR: Efficient DETR Training with Mixed Supervision (CVPR 2024 · similarity 0.87)
- UN-DETR: Promoting Objectness Learning via Joint Supervision for Unknown Object Detection (AAAI 2025 · similarity 0.87)
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 2024 · similarity 0.86)
- Relational Matching for Weakly Semi-Supervised Oriented Object Detection (CVPR 2024 · 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.