Research map / Detection and segmentation / Weakly supervised detection and segmentation
Weakly-Supervised Instance Segmentation via Class-Agnostic Learning With Salient Images
Xinggang Wang, Jiapei Feng, Bin Hu, Qi Ding, Longjin Ran, Xiaoxin Chen, Wenyu Liu
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 |
|---|---|---|---|
| jointly train with salient images | BoxCaseg on PASCAL VOC · PASCAL VOC 2012 val | AP50: 58.9 -> 67.6 | Improved |
| use box-supervised class-agnostic learning with salient images | WSIS on COCO · COCO test-dev | AP: 20.8 -> 30.9 | Improved |
| increase salient training images from 100 to all (ablation) | BoxCaseg · PASCAL VOC 2012 val | mIoU*: 61.8 -> 71.8 | Improved |
| add a weak segmentation head to the salient-only model (ablation) | BoxCaseg · PASCAL VOC 2012 val | mIoU*: 43.1 -> 70.5 | Improved |
| add a transferred segmentation head to the weak-head-only model (ablation) | BoxCaseg · PASCAL VOC 2012 val | mIoU*: 66.1 -> 69.7 | Improved |
| add the salient segmentation head to weak and transferred heads (ablation) | BoxCaseg · PASCAL VOC 2012 val | mIoU*: 69.7 -> 71.8 | Improved |
| use fixed-ratio batch sampling instead of random sampling (ablation) | BoxCaseg · COCO val | mIoU*: 58.1 -> 70.9 | Improved |
| use a 9:7 box-supervised-to-salient batch ratio (ablation) | BoxCaseg · COCO val | mIoU*: 70.7 -> 70.9 | Improved |
| combine salient and transferred segmentation head predictions (ablation) | BoxCaseg · PASCAL VOC 2012 val | mIoU*: 69.7 -> 71.8; 70.5 -> 71.8 | Improved |
| add smaller-object-better merging to proxy masks (ablation) | Proxy Mask R-CNN · COCO val | AP: 28.6 -> 29.5 | Improved |
| add proxy-box-agreement dropping after merging (ablation) | Proxy Mask R-CNN · COCO val | AP: 29.5 -> 30.9 | Improved |
| add proxy-box-agreement dropping after merging (ablation) | Proxy Mask R-CNN · PASCAL VOC 2012 val | AP: 40.2 -> 40.2 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BBAM — BBAM: Bounding Box Attribution Map for Weakly Supervised Semantic and Instance Segmentation (CVPR 2021) · better on 2 of 2 shared results
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
- BoxTeacher: Exploring High-Quality Pseudo Labels for Weakly Supervised Instance Segmentation (CVPR 2023 · similarity 0.90)
- Prior to Segment: Foreground Cues for Weakly Annotated Classes in Partially Supervised Instance Segmentation (ICCV 2021 · similarity 0.90)
- DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence From Box Supervision (ICCV 2021 · similarity 0.89)
- BoxInst: High-Performance Instance Segmentation With Box Annotations (CVPR 2021 · similarity 0.89)
- Box2Mask: Box-Supervised Instance Segmentation via Level-Set Evolution (TPAMI 2024 · similarity 0.89)
- Extreme Point Supervised Instance Segmentation (CVPR 2024 · similarity 0.87)
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.
