Research map / Detection and segmentation / Weakly supervised detection and segmentation
NOTE-RCNN: NOise Tolerant Ensemble RCNN for Semi-Supervised Object Detection
Jiyang Gao, Jiang Wang, Shengyang Dai, Li-Jia Li, Ram Nevatia
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 distillation head to Det-AZ&RPN-A (ablation) | NOTE-RCNN mining · MSCOCO 2017 mined box precision | box precision (%): 83.7 -> 89.9 (iteration 5) | Improved |
| add distillation head to Det-AZ&RPN-A (ablation) | NOTE-RCNN mining · MSCOCO 2017 number of mined boxes | number of mined boxes: 60899 -> 42756 (iteration 5) | Hurt |
| add distillation to NOTE-RCNN (ablation) | NOTE-RCNN · ILSVRC 2013 target categories | mAP: 42.6 -> 43.7 | Improved |
| add seed annotations to FRCN+SemiMIL | Faster RCNN + SemiMIL · ILSVRC 2013 target categories | mAP: 36.9 -> 38.7 | Improved |
| replace Faster RCNN with NOTE-RCNN (ablation) | SemiMIL+Seed · ILSVRC 2013 target categories | mAP: 38.7 -> 39.9 | Improved |
| separate seed and mined annotations with extra heads (vs Naive Faster RCNN) (ablation) | Naive Faster RCNN · MSCOCO 2017 | mAP | Improved |
| zero the background loss in det-cls-a (Det-AZ vs Det-A) (ablation) | Det-A · MSCOCO 2017 and ILSVRC 2013 | mAP | Improved |
| add ensemble classification head to RPN (Det-AZ&RPN-A vs Det-AZ) (ablation) | Det-AZ · MSCOCO 2017 | mAP | Improved |
| add ensemble classification head to RPN (Det-AZ&RPN-A vs Det-AZ) (ablation) | Det-AZ · ILSVRC 2013 | mAP | No clear effect |
| half-distill: distillation in first 4 iterations then none (ablation) | Det-AZ&RPN-A&Distill · MSCOCO 2017 | mAP | Mixed |
| add OpenImage v4 image-level labels with iterative training-mining on full MSCOCO seed | Det-AZ&RPN-A · MSCOCO + OpenImage v4 | mAP@[0.5:0.95]: 32.2 -> 34.0 | Improved |
| vary seed annotation amount (12 to 96 images per category) (ablation) | Det-AZ&RPN-A · MSCOCO 2017 | mAP | Improved |
Closest papers by meaning
- Detecting 11K Classes: Large Scale Object Detection Without Fine-Grained Bounding Boxes (ICCV 2019 · similarity 0.89)
- Unbiased Teacher v2: Semi-Supervised Object Detection for Anchor-Free and Anchor-Based Detectors (CVPR 2022 · similarity 0.87)
- Diverse Learner: Exploring Diverse Supervision for Semi-Supervised Object Detection (ECCV 2022 · similarity 0.87)
- MixTeacher: Mining Promising Labels With Mixed Scale Teacher for Semi-Supervised Object Detection (CVPR 2023 · similarity 0.87)
- Active Learning Strategies for Weakly-Supervised Object Detection (ECCV 2022 · similarity 0.87)
- Multi-clue Consistency Learning to Bridge Gaps Between General and Oriented Object in Semi-supervised Detection (AAAI 2025 · similarity 0.86)
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.
