Research map / Detection and segmentation / Weakly supervised detection and segmentation
Semi-Supervised Object Detection via Virtual Category Learning
Changrui Chen, Kurt Debattista, Jungong Han
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 VC learning instead of discarding confusing samples (ablation) | Unbiased Teacher · MS COCO 1% labels (ablation, small batch) | mAP: 19.37 -> 20.81 | Improved |
| add VC learning instead of keeping all potential labels (ablation) | Unbiased Teacher · MS COCO 1% labels (ablation, small batch) | mAP: 19.36 -> 20.81 | Improved |
| add VC learning to the baseline (ablation) | Unbiased Teacher · MS COCO 1% labels (ablation, small batch) | mAP: 20.00 -> 20.81 | Improved |
| add Reg* loss on top of VC loss (ablation) | VC learning · MS COCO 1% labels (ablation, small batch) | mAP: 20.81 -> 20.94 | Improved |
| use cross-model verification instead of temporal stability to build the PC set (ablation) | VC learning · MS COCO 1% labels (ablation, small batch) | mAP: 20.81 -> 20.96 | Improved |
| add co-training without VC (ablation) | Unbiased Teacher · MS COCO 1% labels (ablation, small batch) | mAP: 20.00 -> 20.53 | Improved |
| flip the input for the virtual weight instead of no augmentation (ablation) | VC learning · MS COCO 1% labels (ablation, small batch) | mAP: 20.80 -> 20.81 | No clear effect |
| use strong augmentation for the virtual weight (ablation) | VC learning · MS COCO 1% labels (ablation, small batch) | mAP: 20.81 -> 20.70 | Hurt |
| set scaling factor 3.5 instead of 2.5 or 4.5 (ablation) | VC learning · MS COCO 1% labels (ablation, small batch) | mAP: 20.63 / 20.65 -> 20.81 | Improved |
| raise location quality threshold t_loc from 0.05 to 0.1 (ablation) | VC learning + Reg* · MS COCO 1% labels (ablation, small batch) | mAP: 20.94 -> 20.31 | Hurt |
| add VC learning plus Soft Teacher tricks | Unbiased Teacher · MS COCO 0.5% labels | mAP: 16.94 -> 19.46 | Improved |
| add VC learning with identical settings | Unbiased Teacher · MS COCO 1% labels | mAP: 20.75 -> 21.61 | Improved |
Closest papers by meaning
- Virtual Category Learning: A Semi-Supervised Learning Method for Dense Prediction With Extremely Limited Labels (TPAMI 2024 · similarity 0.93)
- Diverse Learner: Exploring Diverse Supervision for Semi-Supervised Object Detection (ECCV 2022 · similarity 0.89)
- Rethinking Pseudo Labels for Semi-supervised Object Detection (AAAI 2022 · similarity 0.89)
- Semi-DETR: Semi-Supervised Object Detection With Detection Transformers (CVPR 2023 · similarity 0.89)
- Unbiased Teacher v2: Semi-Supervised Object Detection for Anchor-Free and Anchor-Based Detectors (CVPR 2022 · similarity 0.88)
- Multi-clue Consistency Learning to Bridge Gaps Between General and Oriented Object in Semi-supervised Detection (AAAI 2025 · 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.
