Research map / Detection and segmentation / Weakly supervised detection and segmentation
Affinity Attention Graph Neural Network for Weakly Supervised Semantic Segmentation
Bingfeng Zhang, Jimin Xiao, Jianbo Jiao, Yunchao Wei, Yao Zhao
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 affinity attention layer (replacing random walk) (ablation) | A2GNN · PASCAL VOC 2012 train pseudo labels (box-supervised) | mIoU: 71.3 -> 73.8 | Improved |
| add regularized loss L_reg (ablation) | A2GNN · PASCAL VOC 2012 train pseudo labels (box-supervised) | mIoU: 73.8 -> 74.9 | Improved |
| add multi-point loss L_mp (ablation) | A2GNN · PASCAL VOC 2012 train pseudo labels (box-supervised) | mIoU: 74.9 -> 78.1 | Improved |
| add consistency-checking (ablation) | A2GNN · PASCAL VOC 2012 train pseudo labels (box-supervised) | mIoU: 78.1 -> 78.8 | Improved |
| add multi-point loss to cross-entropy only (ablation) | A2GNN · PASCAL VOC 2012 train pseudo labels (box-supervised) | mIoU: 73.8 -> 75.9 | Improved |
| add consistency-checking to cross-entropy plus L_mp (ablation) | A2GNN · PASCAL VOC 2012 train pseudo labels (box-supervised) | mIoU: 75.9 -> 77.2 | Improved |
| build graph with affinity CNN instead of feature distance (ablation) | A2GNN · PASCAL VOC 2012 train pseudo labels (box-supervised) | mIoU: 74.7 -> 78.8 | Improved |
| use pixels of feature map as nodes instead of superpixels (ablation) | A2GNN · PASCAL VOC 2012 train pseudo labels (box-supervised) | mIoU: 73.3 -> 74.7 | Improved |
| use both cosine attention and edge weights in affinity attention layer vs cosine only (ablation) | A2GNN · PASCAL VOC 2012 train pseudo labels (box-supervised) | mIoU: 73.1 -> 78.8 | Improved |
| use both cosine attention and edge weights vs edge only (ablation) | A2GNN · PASCAL VOC 2012 train pseudo labels (box-supervised) | mIoU: 77.3 -> 78.8 | Improved |
| use confident seed labels M_g instead of M_F as A2GNN supervision (ablation) | A2GNN · PASCAL VOC 2012 train pseudo labels (box-supervised) | mIoU: 73.2 -> 78.8 | Improved |
| add L_Ar to L_Ac when using M_g seeds (ablation) | affinity CNN · PASCAL VOC 2012 train pseudo labels (box-supervised) | mIoU: 71.7 -> 78.8 | Improved |
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.
- Box2Seg — Box2Seg: Attention Weighted Loss and Discriminative Feature Learning for Weakly Supervised Segmentation (ECCV 2020) · 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
- Learning Pixel-Level Semantic Affinity With Image-Level Supervision for Weakly Supervised Semantic Segmentation (CVPR 2018 · similarity 0.90)
- Weakly-Supervised Semantic Segmentation by Iterative Affinity Learning (IJCV 2020 · similarity 0.90)
- Weakly Supervised Semantic Segmentation via Box-Driven Masking and Filling Rate Shifting (TPAMI 2023 · similarity 0.90)
- Learning Affinity From Attention: End-to-End Weakly-Supervised Semantic Segmentation With Transformers (CVPR 2022 · similarity 0.89)
- Leveraging Auxiliary Tasks With Affinity Learning for Weakly Supervised Semantic Segmentation (ICCV 2021 · similarity 0.88)
- Box2Seg: Attention Weighted Loss and Discriminative Feature Learning for Weakly Supervised Segmentation (ECCV 2020 · 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.
