Research map / Detection and segmentation / Weakly supervised detection and segmentation
Progressive Representation Adaptation for Weakly Supervised Object Localization
Dong Li, Jia‐Bin Huang, Yali Li, Shengjin Wang, Ming–Hsuan Yang
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 MIL selection to class-specific proposals (ablation) | class-specific proposals · VOC 2007 trainval | CorLoc: 31.8 -> 41.2 | Improved |
| add detection network fine-tuning (ablation) | CS+MIL · VOC 2007 trainval | CorLoc: 41.2 -> 49.8 (AlexNet), 52.4 (VGGNet) | Improved |
| add activation scores (ablation) | CS+MIL+FT (VGGNet) · VOC 2007 trainval | CorLoc: 52.4 -> 54.5 | Improved |
| add segmentation cues for box refinement (ablation) | CS+AS+MIL+FT · VOC 2007 trainval | CorLoc: 54.5 -> 59.0 | Improved |
| train detector on MIL-selected proposals directly (no classification adaptation) (ablation) | MIL baseline (AlexNet) · VOC 2007 test | mAP: 22.7 -> 23.0 | No clear effect |
| add class-specific proposal mining from adapted classifier (ablation) | MIL+FT (AlexNet) · VOC 2007 test | mAP: 23.0 -> 31.0 | Improved |
| add class-specific proposal mining from adapted classifier (ablation) | MIL+FT (VGGNet) · VOC 2007 test | mAP: 26.2 -> 39.5 | Improved |
| add MIL step to class-specific proposals (ablation) | CS+FT (AlexNet / VGGNet) · VOC 2007 test | mAP: 19.5 -> 31.0 (AlexNet), 20.5 -> 39.5 (VGGNet) | Improved |
| add activation scores (ablation) | CS+MIL+FT (VGGNet) · VOC 2007 test | mAP: 39.5 -> 41.2 | Improved |
| add proposal refinement with segmentation (ablation) | CS+AS+MIL+FT (VGGNet) · VOC 2007 test | mAP: 41.2 -> 42.5 | Improved |
| use VGGNet instead of AlexNet | CS+MIL+FT · VOC 2007 test | mAP: 31.0 -> 39.5 | Improved |
| use In-Out mask-out instead of Whole-Out (ablation) | proposal mining · VOC 2007 train/val | CorLoc@M: 64.9 -> 73.8 (M=10) | Improved |
Closest papers by meaning
- Weakly Supervised Object Localization With Progressive Domain Adaptation (CVPR 2016 · similarity 0.97)
- Weakly Supervised Object Localization and Detection: A Survey (TPAMI 2021 · similarity 0.90)
- You Reap What You Sow: Using Videos to Generate High Precision Object Proposals for Weakly-Supervised Object Detection (CVPR 2019 · similarity 0.89)
- PreDet: Large-Scale Weakly Supervised Pre-Training for Detection (ICCV 2021 · similarity 0.89)
- Pairwise Similarity Knowledge Transfer for Weakly Supervised Object Localization (ECCV 2020 · similarity 0.89)
- Soft Proposal Networks for Weakly Supervised Object Localization (ICCV 2017 · similarity 0.89)
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.
