Research map / Detection and segmentation / Weakly supervised detection and segmentation
Open-Set Semi-Supervised Object Detection
Yen-Cheng Liu, Chih-Yao Ma, Xiaoliang Dai, Junjiao Tian, Peter Vajda, Zijian He, Zsolt Kira
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 DINO-based offline OOD filtering to pseudo-labels | Unbiased Teacher · COCO-Open (20 ID classes, 4k labeled) | mAP: 18.37 -> 23.43 | Improved |
| add OOD filtering | Unbiased Teacher · COCO-Open (40 ID classes, 4k labeled) | mAP: 20.28 -> 22.91 | Improved |
| add OOD filtering | Unbiased Teacher · COCO-Open (60 ID classes, 4k labeled) | mAP: 23.09 -> 24.89 | Improved |
| add OOD filtering with only 1,000 labeled images | Unbiased Teacher · COCO-Open (20 ID classes) | mAP: 11.77 -> 16.80 | Improved |
| add OOD filtering with 4,000 labeled images | Unbiased Teacher · COCO-Open (20 ID classes) | mAP: 18.23 -> 22.56 | Improved |
| add mixed and pure-OOD images to unlabeled set (no filtering) | Unbiased Teacher · COCO-Open (40 ID classes) | mAP: 22.10 -> lower | Hurt |
| use offline DINO OOD detector instead of online ROIhead detector (Euclidean vs MSP) (ablation) | OOD detector · OOD detection, COCO 20 ID classes | mAP: AUROC 67.0 -> 90.8 | Improved |
| use Euclidean distance instead of Mahalanobis for offline OOD score (ablation) | DINO OOD detector · OOD detection, COCO 20 ID classes | mAP: AUROC 81.8 -> 90.8 | Improved |
| add OOD filtering for large-scale OpenImages unlabeled data | Unbiased Teacher · COCO-OpenImages | mAP: 41.81 -> 43.14 | Improved |
| add OOD filtering on COCO-unlabeled | Unbiased Teacher* · COCO-additional | mAP: 44.06 -> 45.14 | Improved |
| add OOD filtering to STAC | STAC · COCO-Open (40 ID classes, 4k labeled) | mAP: 18.60 -> 19.80 | Improved |
| use supervised ViT-B OOD detector instead of DINO (ablation) | UT + OOD filtering · COCO-Open (20 ID classes, 1,000 labeled images) | mAP: 16.80 -> 17.10 | Improved |
Closest papers by meaning
- Semi-supervised Open-World Object Detection (AAAI 2024 · similarity 0.91)
- Unbiased Teacher v2: Semi-Supervised Object Detection for Anchor-Free and Anchor-Based Detectors (CVPR 2022 · similarity 0.91)
- Semi-DETR: Semi-Supervised Object Detection With Detection Transformers (CVPR 2023 · similarity 0.91)
- Toward Efficient Semi-Supervised Object Detection With Detection Transformer (TPAMI 2025 · similarity 0.90)
- Unbiased Teacher for Semi-Supervised Object Detection (ICLR 2021 · similarity 0.90)
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 2024 · similarity 0.90)
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.
