Research map / Detection and segmentation / CNN features
Squeeze-and-Attention Networks for Semantic Segmentation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace the FCN prediction head with four SA heads (ablation) | ResNet50 FCN · PASCAL Context (59 classes without background) | mIoU: 43.2 -> 49.2 | Helped |
| add categorical loss (ablation) | ResNet50 SANet with SA heads · PASCAL Context (59 classes without background) | mIoU: 49.2 -> 50.7 | Helped |
| add dense prediction loss (ablation) | ResNet50 SANet with SA heads and categorical loss · PASCAL Context (59 classes without background) | mIoU: 50.7 -> 51.9 | Helped |
| use ResNet101 rather than ResNet50 as the backbone (ablation) | SANet with all losses · PASCAL Context (59 classes without background) | mIoU: 51.9 -> 53.0 | Helped |
| use EfficientNet-b7 rather than ResNet101 as the backbone (ablation) | SANet · PASCAL Context (59 classes without background) | mIoU: 53.0 -> 55.3 | Helped |
| add SE modules to the FCN (ablation) | FCN50 · PASCAL Context | mIoU: 44.9 -> 44.6 | Hurt |
| add SE modules to the FCN (ablation) | FCN101 · PASCAL Context | mIoU: 45.6 -> 45.7 | No clear effect |
| use SANet instead of FCN | ResNet50 segmentation model · PASCAL Context | mIoU: 44.9 -> 49.0 | Helped |
| use SANet instead of FCN | ResNet101 segmentation model · PASCAL Context | mIoU: 45.6 -> 50.1 | Helped |
| use SANet instead of Deeplab V3+ | ResNet101 segmentation model · PASCAL VOC test without COCO pretraining | mIoU: 81.5 -> 83.2 | Helped |
Closest papers by meaning
- Attention to Scale: Scale-Aware Semantic Image Segmentation (CVPR 2016 · similarity 0.90)
- Fully Attentional Network for Semantic Segmentation (AAAI 2022 · similarity 0.88)
- RANet: Region Attention Network for Semantic Segmentation (NeurIPS 2020 · similarity 0.88)
- AttaNet: Attention-Augmented Network for Fast and Accurate Scene Parsing (AAAI 2021 · similarity 0.88)
- Unveiling the Knowledge of CLIP for Training-Free Open-Vocabulary Semantic Segmentation (AAAI 2025 · similarity 0.86)
- Revisiting the Integration of Convolution and Attention for Vision Backbone (NeurIPS 2024 · similarity 0.86)
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.