Research map / Detection and segmentation / Weakly supervised detection and segmentation
STC: A Simple to Complex Framework for Weakly-Supervised Semantic Segmentation
Yunchao Wei, Xiaodan Liang, Yunpeng Chen, Xiaohui Shen, Ming‐Ming Cheng, Jiashi Feng, Yao Zhao, Shuicheng Yan
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 simple-to-complex stage E-DCNN (train on I-DCNN predicted masks) (ablation) | I-DCNN · VOC 2012 val | mIoU: 44.9 -> 46.3 | Improved |
| add complex images (P-DCNN stage) on top of E-DCNN (ablation) | E-DCNN · VOC 2012 val | mIoU: 46.3 -> 49.8 | Improved |
| use DRFI saliency maps instead of HS (ablation) | I-DCNN · VOC 2012 val | mIoU: 42.5 -> 44.9 | Improved |
| use simple images instead of complex images to train I-DCNN (ablation) | I-DCNN · VOC 2012 val | mIoU: 17.6 -> 44.9 | Improved |
| use multi-label cross-entropy instead of SaliencyCut masks with single-label loss (ablation) | I-DCNN · VOC 2012 val | mIoU: drop by 3% | Improved |
| filter crawled images to simple ones (0.3*N_I<N_f<0.5*N_I) (ablation) | I-DCNN · VOC 2012 val | mIoU: 4% performance drop | Improved |
| increase Flickr-Clean images from 1/16 to 1/4 (ablation) | I-DCNN · VOC 2012 val | mIoU: 39.8 -> 45.7 | Improved |
| increase Flickr-Clean images from 1/4 to All (ablation) | I-DCNN · VOC 2012 val | mIoU: 45.7 -> 44.9 | Hurt |
| use cross-entropy only vs SEC constrain-to-boundary loss | STC vs SEC · VOC 2012 val | mIoU: 49.8 vs. 50.7 | No clear effect |
Closest papers by meaning
- Combining Bottom-Up, Top-Down, and Smoothness Cues for Weakly Supervised Image Segmentation (CVPR 2017 · similarity 0.91)
- Weakly-Supervised Semantic Segmentation by Iteratively Mining Common Object Features (CVPR 2018 · similarity 0.91)
- Webly Supervised Semantic Segmentation (CVPR 2017 · similarity 0.90)
- Seed, Expand and Constrain: Three Principles for Weakly-Supervised Image Segmentation (ECCV 2016 · similarity 0.90)
- Single-Stage Semantic Segmentation From Image Labels (CVPR 2020 · similarity 0.90)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 2024 · 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.
