Research map / Detection and segmentation / Weakly supervised detection and segmentation
Soft Proposal Networks for Weakly Supervised Object Localization
Yi Zhu, Yanzhao Zhou, Qixiang Ye, Qiang Qiu, Jianbin Jiao
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 |
|---|---|---|---|
| upgrade VGG16 to SPN by adding Soft Proposal module | VGG16 · VOC2007 test pointing (All) | pointing accuracy: 80.0 -> 87.5 | Improved |
| upgrade VGG16 to SPN by adding Soft Proposal module | VGG16 · VOC2007 test pointing (Diff.) | pointing accuracy: 66.8 -> 78.1 | Improved |
| upgrade GoogLeNet to SPN by adding Soft Proposal module | GoogLeNet · VOC2007 test pointing (All/Diff.) | pointing accuracy: 85.1/72.3 -> 88.2/79.1 | Improved |
| upgrade CNN-S to SPN | CNN-S · VOC2007 test pointing (All/Diff.) | pointing accuracy: 78.7/61.7 -> 81.8/66.7 | Improved |
| use weakly supervised soft proposal instead of supervised RPN | SPN · VOC2007 test proposal quality | Object Energy: 63.3 -> 62.2 | No clear effect |
| use soft proposal instead of RPN for proposal time | SPN · VOC2007 test proposal quality | time per image (ms): 10.5 -> 0.9 | Improved |
| use soft proposal instead of Selective Search | SPN · VOC2007 test proposal quality | Object Energy: 53.7 -> 62.2 | Improved |
| upgrade pre-trained VGG16 to SPN, pointing with prediction | VGG16 · VOC2012 val pointing with prediction | mAP: 77.1 -> 82.9 | Improved |
| upgrade pre-trained VGG16 to SPN, pointing with prediction | VGG16 · COCO2014 val pointing with prediction | mAP: 49.2 -> 55.3 | Improved |
| use SP-VGGNet for bounding box localization | VGG16 · VOC2007 trainval positive images CorLoc | CorLoc mean: 55.1 -> 60.6 | Improved |
| add Soft Proposal module to GoogLeNetGAP | GoogLeNetGAP · ILSVRC2014 val classification | top-1/top-5 error: 35.0/13.2 -> 33.5/12.7 | Improved |
| add Soft Proposal module to GoogLeNetGAP then fine-tune | GoogLeNetGAP · VOC2007 test classification (fine-tuned) | mAP: 83.4 -> 84.2 | Improved |
Closest papers by meaning
- Weakly Supervised Localization Using Deep Feature Maps (ECCV 2016 · similarity 0.91)
- 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 0.90)
- Two-Phase Learning for Weakly Supervised Object Localization (ICCV 2017 · similarity 0.89)
- Weakly Supervised Object Localization With Progressive Domain Adaptation (CVPR 2016 · similarity 0.89)
- Weakly Supervised Cascaded Convolutional Networks (CVPR 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.
