Research map / Detection and segmentation / Weakly supervised detection and segmentation
C-MIDN: Coupled Multiple Instance Detection Network With Segmentation Guidance for Weakly Supervised Object Detection
Yan Gao, Boxiao Liu, Nan Guo, Xiaochun Ye, Fang Wan, Haihang You, Dongrui Fan
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 |
|---|---|---|---|
| replace single WSDDN with coupled pair of MIDNs (C-MIDN) with SGPR (ablation) | WSDDN+ODR baseline · PASCAL VOC 2007 test | mAP: 49.0 -> 52.6 | Improved |
| replace single WSDDN with C-MIDN | WSDDN+ODR baseline · PASCAL VOC 2012 test | mAP: 46.8 -> 50.2 | Improved |
| add Fast R-CNN retraining on C-MIDN pseudo ground truth | C-MIDN · PASCAL VOC 2007 test | mAP: 52.6 -> 53.6 | Improved |
| add Fast R-CNN retraining on C-MIDN pseudo ground truth | C-MIDN · PASCAL VOC 2012 test | mAP: 50.2 -> 50.3 | No clear effect |
| use C-MIDN instead of baseline for CorLoc (ablation) | WSDDN+ODR baseline · PASCAL VOC 2007 trainval | CorLoc | Improved |
| add Fast R-CNN retraining | C-MIDN · PASCAL VOC 2007 trainval | CorLoc: 68.7 -> 71.9 | Improved |
| use segmentation guided proposal removal instead of unconditional proposal removal (ablation) | C-MIDN · PASCAL VOC 2007 test | mAP: UPR performance declines greatly vs SGPR | Improved |
| vary IoU removal threshold t_remove (ablation) | C-MIDN · PASCAL VOC 2007 test | mAP: all thresholds outperform baseline by more than 2.4 mAP; peak at 0.3 | No clear effect |
| add C-MIDN on cat class (ablation) | WSDDN+ODR baseline · PASCAL VOC 2007 test | AP cat: 47 -> 68.3 | Improved |
| add C-MIDN on person class (ablation) | WSDDN+ODR baseline · PASCAL VOC 2007 test | AP person: 7 -> 20 | Improved |
| add C-MIDN on dog class (ablation) | WSDDN+ODR baseline · PASCAL VOC 2007 test | AP dog: 59.8 -> 64.6 | Improved |
| add C-MIDN on diningtable and aeroplane classes (ablation) | WSDDN+ODR baseline · PASCAL VOC 2007 test | AP: 37.5 -> 45.1 (diningtable), 44.3 -> 53.3 (aeroplane) | Improved |
Later papers that compared with this method
2 later accepted papers report C-MIDN as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Instance-Aware, Context-Focused, and Memory-Efficient Weakly Supervised Object Detection (CVPR 2020 · better on 1 of 1 shared result)
- UFO²: A Unified Framework towards Omni-supervised Object Detection (ECCV 2020 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Cyclic-Bootstrap Labeling for Weakly Supervised Object Detection (ICCV 2023 · similarity 0.90)
- PCL: Proposal Cluster Learning for Weakly Supervised Object Detection (TPAMI 2018 · similarity 0.89)
- Towards Precise End-to-End Weakly Supervised Object Detection Network (ICCV 2019 · similarity 0.89)
- C-MIL: Continuation Multiple Instance Learning for Weakly Supervised Object Detection (CVPR 2019 · similarity 0.89)
- Weakly Supervised Object Detection With Segmentation Collaboration (ICCV 2019 · similarity 0.88)
- Multiple Instance Detection Network With Online Instance Classifier Refinement (CVPR 2017 · 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.
