Research map / Detection and segmentation / CNN features
PSANet: Point-wise Spatial Attention Network for Scene Parsing
Hengshuang Zhao, Yi Zhang, Shu Liu, Jianping Shi, Chen Change Loy, Dahua Lin, Jiaya Jia
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 COLLECT branch (ablation) | ResNet-50 FCN baseline · ADE20K val (SS) | mIoU (%): 37.23 -> 41.27 | Improved |
| add DISTRIBUTE branch to COLLECT (bi-direction) (ablation) | PSANet ResNet-50 · ADE20K val (SS) | mIoU (%): 41.27 -> 41.92 | Improved |
| add DISTRIBUTE branch to COLLECT (bi-direction) (ablation) | PSANet ResNet-101 · ADE20K val (SS) | mIoU (%): 42.70 -> 42.75 | Improved |
| use over-complete masks with relative location instead of compact masks (ablation) | PSA COLLECT ResNet-50 · ADE20K val (SS) | mIoU (%): 41.07 -> 41.27 | Improved |
| replace non-local correlation masks with learned compact PSA masks (ablation) | ResNet-50 · ADE20K val (SS) | mIoU (%): 40.93 -> 41.07 | Improved |
| add PSA module (COLLECT+DISTRIBUTE) instead of PSP pooling | ResNet-50 · ADE20K val (SS) | mIoU (%): 41.68 -> 41.92 | Improved |
| add DenseCRF post-processing | ResNet-50 baseline · ADE20K val (SS) | mIoU (%): 37.23 -> 37.97 | No clear effect |
| add PSA module (COLLECT+DISTRIBUTE) (ablation) | ResNet-50 · VOC 2012 val (SS) | mIoU (%): 67.12 -> 77.24 | Improved |
| add PSA module (COLLECT+DISTRIBUTE) (ablation) | ResNet-101 · VOC 2012 val (SS) | mIoU (%): 70.64 -> 78.51 | Improved |
| add DISTRIBUTE to COLLECT (ablation) | ResNet-101 · VOC 2012 val (SS) | mIoU (%): 77.90 -> 78.51 | Improved |
| add PSA module (COLLECT+DISTRIBUTE) (ablation) | ResNet-50 · Cityscapes val (SS) | mIoU (%): 71.93 -> 76.65 | Improved |
| add PSA module (COLLECT+DISTRIBUTE) (ablation) | ResNet-101 · Cityscapes val (SS) | mIoU (%): 74.83 -> 77.94 | Improved |
Later papers that compared with this method
14 later accepted papers report PSANet50 as a baseline in their results tables; 14 of them report a better number on most of the results they share with it.
- AutoBSS: An Efficient Algorithm for Block Stacking Style Search (NeurIPS 2020 · better on 1 of 1 shared result)
- Multiscale Deep Equilibrium Models (NeurIPS 2020 · better on 1 of 1 shared result)
- Class-wise Dynamic Graph Convolution for Semantic Segmentation (ECCV 2020 · better on 1 of 1 shared result)
- Expectation-Maximization Attention Networks for Semantic Segmentation (ICCV 2019 · better on 3 of 3 shared results)
- ACFNet: Attentional Class Feature Network for Semantic Segmentation (ICCV 2019 · better on 2 of 2 shared results)
- Learnable Tree Filter for Structure-preserving Feature Transform (NeurIPS 2019 · better on 1 of 1 shared result)
- Asymmetric Non-Local Neural Networks for Semantic Segmentation (ICCV 2019 · better on 1 of 1 shared result)
- CCNet: Criss-Cross Attention for Semantic Segmentation (ICCV 2019 · better on 1 of 1 shared result)
6 more not listed.
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
- Perspective-Adaptive Convolutions for Scene Parsing (TPAMI 2019 · similarity 0.88)
- FoveaNet: Perspective-Aware Urban Scene Parsing (ICCV 2017 · similarity 0.88)
- Adaptive Context Network for Scene Parsing (ICCV 2019 · similarity 0.87)
- Pyramid Scene Parsing Network (CVPR 2017 · similarity 0.87)
- Strip Pooling: Rethinking Spatial Pooling for Scene Parsing (CVPR 2020 · similarity 0.87)
- AttaNet: Attention-Augmented Network for Fast and Accurate Scene Parsing (AAAI 2021 · 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.
