Research map / Detection and segmentation / Weakly supervised detection and segmentation
Semi-supervised Object Detection with Adaptive Class-Rebalancing Self-Training
Read the paperCheck 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 |
|---|---|---|---|
| use ACRST with CropBank-based rebalancing and pseudo-label filtering | FPN-Faster-RCNN, ResNet-50 · COCO-standard 1% | AP50:95: 20.75 -> 26.07 | Helped |
| use ACRST with CropBank-based rebalancing and pseudo-label filtering | FPN-Faster-RCNN, ResNet-50 · COCO-additional | AP50:95: 41.30 -> 42.79 | Helped |
| use ACRST with CropBank-based rebalancing and pseudo-label filtering | FPN-Faster-RCNN, ResNet-50 · VOC07-test (VOC07&12) | AP50:95: 48.69 -> 50.12 | Helped |
| add foreground-background rebalancing (ablation) | Mean Teacher · COCO-standard 1% | AP50:95: 20.75 -> 23.32 | Helped |
| add adaptive foreground-foreground rebalancing (ablation) | Mean Teacher + FBR · COCO-standard 1% | AP50:95: 23.32 -> 24.17 | Helped |
| add two-stage pseudo-label filtering (ablation) | Mean Teacher · COCO-standard 1% | AP50:95: 20.75 -> 23.48 | Helped |
| add two-stage pseudo-label filtering (ablation) | Mean Teacher + FBR + AFFR · COCO-standard 1% | AP50:95: 24.17 -> 25.56 | Helped |
| add selective supervision for regression on CropBank instances (ablation) | ACRST · COCO-standard 1% | AP50:95: 25.56 -> 26.12 | Helped |
| replace MixUp and Mosaic with CropBank augmentation (ablation) | STAC · COCO-standard 1% | AP50:95: 16.00 -> 16.85 | Helped |
| use two-stage mining rather than confidence-only filtering (ablation) | One-stage pseudo-labeling · COCO-standard 1% | pseudo-label recall: 0.299 -> 0.376 | Helped |
| use two-stage mining rather than confidence-only filtering (ablation) | One-stage pseudo-labeling · COCO-standard 1% | pseudo-label accuracy: 0.735 -> 0.722 | Hurt |
| filter with image-level semantics after confidence thresholding (ablation) | One-stage pseudo-label filtering · COCO-standard 1% | pseudo-label accuracy: 0.735 -> 0.792 | Helped |
Closest papers by meaning
- Adapting Object Size Variance and Class Imbalance for Semi-supervised Object Detection (AAAI 2023 · similarity 0.92)
- Rethinking Pseudo Labels for Semi-supervised Object Detection (AAAI 2022 · similarity 0.91)
- Instant-Teaching: An End-to-End Semi-Supervised Object Detection Framework (CVPR 2021 · similarity 0.91)
- Unbiased Teacher for Semi-Supervised Object Detection (ICLR 2021 · similarity 0.90)
- Consistent-Teacher: Towards Reducing Inconsistent Pseudo-Targets in Semi-Supervised Object Detection (CVPR 2023 · similarity 0.89)
- 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.