Research map / Detection and segmentation / Weakly supervised detection and segmentation
Instant-Teaching: An End-to-End Semi-Supervised Object Detection Framework
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 |
|---|---|---|---|
| Use end-to-end instant pseudo-label generation instead of STAC's pre-generated, fixed pseudo labels | Instant-Teaching · MS-COCO 2% labeled | mAP: 18.25±0.25 -> 20.70±0.30 | Helped |
| Add co-rectify using a second independently initialized model (ablation) | Instant-Teaching · MS-COCO 2% labeled | mAP: 20.70±0.30 -> 22.45±0.15 | Helped |
| Add Mixup (ablation) | Instant-Teaching with Color+Cutout · MS-COCO 5% labeled, 8× unlabeled | mAP: 24.70 -> 25.40 | Helped |
| Add Mosaic (ablation) | Instant-Teaching with Color+Cutout · MS-COCO 5% labeled, 8× unlabeled | mAP: 24.70 -> 25.00 | Helped |
| Add both Mixup and Mosaic (ablation) | Instant-Teaching with Color+Cutout · MS-COCO 5% labeled, 8× unlabeled | mAP: 24.70 -> 25.60 | Helped |
| Apply Color+Cutout strong augmentation during pseudo-label generation (ablation) | Instant-Teaching · MS-COCO 5% labeled, 8× unlabeled | mAP: 23.14 -> 21.60 | Hurt |
| Increase confidence threshold from 0.3 to 0.9 (ablation) | Instant-Teaching · MS-COCO 10% labeled | mAP: 26.30 -> 29.80 | Helped |
| Increase the unlabeled-data scale from 1× to 8× (ablation) | Instant-Teaching · MS-COCO 5% labeled | mAP: 23.60 -> 25.60 | Helped |
| Use VOC12 and additional COCO images as unlabeled data rather than VOC12 alone | Instant-Teaching* · VOC07 test | AP@0.5:0.95: 50.00 -> 50.80 | Helped |
| Replace R50-FPN with a more powerful backbone (ablation) | Instant-Teaching* · MS-COCO 2% labeled | mAP: 22.45±0.15 -> 23.50±0.20 (R101-FPN); 24.20±0.15 (X101-32x4d-FPN); 25.9±0.20 (R2N-101-FPN) | Helped |
| Use a smaller unsupervised-loss weight early in training (ablation) | Instant-Teaching · MS-COCO 10% labeled | mAP | Helped |
Closest papers by meaning
- Unbiased Teacher for Semi-Supervised Object Detection (ICLR 2021 · similarity 0.94)
- Humble Teachers Teach Better Students for Semi-Supervised Object Detection (CVPR 2021 · similarity 0.93)
- Active Teacher for Semi-Supervised Object Detection (CVPR 2022 · similarity 0.92)
- Point-Teaching: Weakly Semi-supervised Object Detection with Point Annotations (AAAI 2023 · similarity 0.91)
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 2024 · similarity 0.88)
- Relational Matching for Weakly Semi-Supervised Oriented Object Detection (CVPR 2024 · similarity 0.88)
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.