Research map / Detection and segmentation / CNN features
Supervised Edge Attention Network for Accurate Image Instance Segmentation
Xier Chen, Yanchao Lian, Licheng Jiao, Haoran Wang, YanJie Gao, Shi Lingling
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 SEANet (box head, B-IoU, edge attention) | Mask R-CNN ResNet-101-FPN · COCO val | mask AP: 36.5 -> 37.7 | Improved |
| add SEANet | Mask R-CNN ResNet-50-FPN · COCO val | mask AP: 34.5 -> 36.0 | Improved |
| add SEANet | Mask R-CNN ResNeXt-101-FPN · COCO val | mask AP: 37.7 -> 39.4 | Improved |
| add SEANet | MS R-CNN ResNet-101-FPN · COCO val | mask AP: 37.8 -> 38.5 | Improved |
| add SEANet | MS R-CNN ResNeXt-101-FPN · COCO val | box AP: 42.1 -> 43.7 | Improved |
| add SEANet | Mask R-CNN ResNeXt-101-FPN · COCO test-dev | mask AP: 38.2 -> 39.7 | Improved |
| add edge supervision to the attention module (ablation) | SEANet mask head · COCO val R50 | mask AP: 35.1 -> 36.0 | Improved |
| fuse edge attention by sum instead of concat (ablation) | SEANet mask head · COCO val R50 | mask AP: 35.6 -> 36.0 | Improved |
| add SEANet to improve boundary quality | Mask R-CNN · COCO val band width 8 | boundary accuracy: 41.47 -> 57.21 | Improved |
| add SEANet | Mask R-CNN ResNet-101-FPN · runtime | runtime ms: 405 -> 464 | Hurt |
| add SEANet | Mask R-CNN ResNet-101-FPN · parameters | params M: 63.17 -> 64.89 | No clear effect |
Closest papers by meaning
- BoxInst: High-Performance Instance Segmentation With Box Annotations (CVPR 2021 · similarity 0.88)
- The Surprising Impact of Mask-Head Architecture on Novel Class Segmentation (ICCV 2021 · similarity 0.87)
- Learning to Segment Every Thing (CVPR 2018 · similarity 0.87)
- Box2Mask: Box-Supervised Instance Segmentation via Level-Set Evolution (TPAMI 2024 · similarity 0.87)
- Pseudo Mask Augmented Object Detection (CVPR 2018 · similarity 0.87)
- Extreme Point Supervised Instance Segmentation (CVPR 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.
