Research map / Detection and segmentation / Weakly supervised detection and segmentation
Scale-Equivalent Distillation for Semi-Supervised Object Detection
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 |
|---|---|---|---|
| Apply SED rather than supervised-only training | Faster R-CNN · MS-COCO 2017 val, 5% labeled | mAP: 18.47 -> 29.01 | Helped |
| Apply SED rather than supervised-only training | Faster R-CNN · MS-COCO 2017 val, 10% labeled | mAP: 23.86 -> 34.02 | Helped |
| Apply SED rather than the strongest reported semi-supervised baseline | Faster R-CNN · Pascal VOC 2007 test | AP50: 78.60 -> 80.60 | Helped |
| Add scale consistency regularization (ablation) | SED · MS-COCO 2017 val, 10% labeled | mAP: 23.86 -> 26.80 | Helped |
| Add reweighting to scale consistency regularization (ablation) | SED with SCR · MS-COCO 2017 val, 10% labeled | mAP: 26.80 -> 30.10 | Helped |
| Add soft self-distillation (ablation) | SED · MS-COCO 2017 val, 10% labeled | mAP: 23.86 -> 29.80 | Helped |
| Add reweighting to soft self-distillation (ablation) | SED with soft self-distillation · MS-COCO 2017 val, 10% labeled | mAP: 29.80 -> 31.40 | Helped |
| Use soft rather than hard self-distillation targets (ablation) | SED with SCR and reweighting · MS-COCO 2017 val, 10% labeled | mAP: 29.50 -> 34.00 | Helped |
| Use SCR, soft self-distillation, and reweighting together (ablation) | SED · MS-COCO 2017 val, 10% labeled | mAP: 23.86 -> 34.00 | Helped |
| Set SCR downsampling rate to 2 rather than rate 1 (ablation) | SED with RetinaNet and ResNet-18 · MS-COCO val | mAP: 23.0 -> 26.1 | Helped |
| Increase SCR downsampling rate from 2 to 16 (ablation) | SED with RetinaNet and ResNet-18 · MS-COCO val | mAP: 26.1 -> 21.1 | Hurt |
| Use step-decayed teacher EMA instead of fixed EMA (ablation) | SED · MS-COCO val | mAP: 32.0 -> 34.1 | Helped |
Closest papers by meaning
- Unbiased Teacher v2: Semi-Supervised Object Detection for Anchor-Free and Anchor-Based Detectors (CVPR 2022 · similarity 0.90)
- Unbiased Teacher for Semi-Supervised Object Detection (ICLR 2021 · similarity 0.90)
- Adapting Object Size Variance and Class Imbalance for Semi-supervised Object Detection (AAAI 2023 · similarity 0.90)
- ScaleKD: Distilling Scale-Aware Knowledge in Small Object Detector (CVPR 2023 · similarity 0.90)
- SimLTD: Simple Supervised and Semi-Supervised Long-Tailed Object Detection (CVPR 2025 · similarity 0.87)
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 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.