Research map / Detection and segmentation / Weakly supervised detection and segmentation
Boosting Weakly Supervised Object Detection with Progressive Knowledge Transfer
Yuanyi Zhong, Jianfeng Wang, Jian Peng, Lei Zhang
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 5 progressive refinements of OCUD and MIL classifier (ablation) | Ours · VOC 2007 test | mAP: 55.0 -> 59.7 | Improved |
| use multi-step instead of one-step transfer (ablation) | Ours · VOC 2007 test (COCO-60) | mAP: 54.93 -> 59.71 | Improved |
| add progressive refinements under ILSVRC transfer (ablation) | Ours · ILSVRC2013 val2 weak 100 | mAP: 33.5 -> 37.0 | Improved |
| add multi-scale testing with flipping | Ours · VOC 2007 test | mAP: 57.77 -> 59.71 | Improved |
| add multi-scale testing with flipping | Ours · VOC 2007 trainval | CorLoc: 73.6 -> 74.4 | Improved |
| retrain Faster RCNN on mined pseudo ground truth (distill, ResNet50) | Ours+ · VOC 2007 test | mAP: 59.71 -> 60.24 | Improved |
| remove source dataset during refinements (VOC only) (ablation) | Ours · VOC 2007 test | mAP | Hurt |
| use COCO-60 instead of COCO-60-full as source (ablation) | Ours · VOC 2007 test | mAP | Improved |
| use COCO-60 instead of ILSVRC-179 as source (ablation) | Ours · VOC 2007 test | mAP: 56.46 -> 59.7 | Improved |
| set pseudo GT confidence threshold tau to 0.8 versus lower values (ablation) | Ours · VOC 2007 test | mAP | Mixed |
| use detection score scale beta = 5 (ablation) | MIL classifier · VOC 2007 test | mAP | Improved |
| add guidance regularization lambda = 0.2 from OCUD (ablation) | MIL classifier · VOC 2007 test | mAP | Improved |
Closest papers by meaning
- Revisiting Knowledge Transfer for Training Object Class Detectors (CVPR 2018 · similarity 0.93)
- Weak-Shot Object Detection Through Mutual Knowledge Transfer (CVPR 2023 · similarity 0.92)
- CaT: Weakly Supervised Object Detection With Category Transfer (ICCV 2021 · similarity 0.91)
- W2F: A Weakly-Supervised to Fully-Supervised Framework for Object Detection (CVPR 2018 · similarity 0.91)
- Weakly Supervised Knowledge Transfer with Probabilistic Logical Reasoning for Object Detection (ICLR 2023 · similarity 0.90)
- Cross-Domain Weakly-Supervised Object Detection Through Progressive Domain Adaptation (CVPR 2018 · similarity 0.90)
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.
