Research map / Detection and segmentation / CNN features
ELSD: Efficient Line Segment Detector and Descriptor
Haotian Zhang, Yicheng Luo, Fangbo Qin, Yijia He, Xiao Liu
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 non-centerness suppression (ablation) | ELSD · Wireframe | sAP10: 68.0 -> 68.9 | Improved |
| add descriptor branch with joint training (ablation) | ELSD · Wireframe | sAP10: 68.9 -> 68.5 | No clear effect |
| upsample prediction maps to 256 resolution (ablation) | ELSD · Wireframe | sAP10: 65.8 -> 68.9 | Improved |
| use focal loss instead of BCE for midpoint map (ablation) | ELSD · Wireframe | sAP10: 66.0 -> 68.9 | Improved |
| use CAL representation instead of Tri-points (ablation) | ELSD · Wireframe | sAP10: 67.9 -> 68.9 | Improved |
| all four components together versus none (ablation) | ELSD · Wireframe | sAP10: 62.8 -> 68.9 | Improved |
| use 64-d descriptor instead of 36-d (ablation) | ELSD · Wireframe | F-score: 73.7 -> 74.8 | Improved |
| use 64-d descriptor instead of 256-d | ELSD · Line matching (ScanNet subset) | F-score: 74.7 -> 74.8 | No clear effect |
| replace ResNet34 backbone with Hourglass | ELSD · Wireframe | sAP10: 68.9 -> 67.2 | Hurt |
| shrink to Lite version (256 input) | ELSD · Wireframe | sAP10: 68.9 -> 63.1 | Hurt |
Closest papers by meaning
- Unified and Efficient Point-Line Local Features (ECCV 2026 · similarity 0.87)
- ScaleLSD: Scalable Deep Line Segment Detection Streamlined (CVPR 2025 · similarity 0.87)
- Towards Light-Weight and Real-Time Line Segment Detection (AAAI 2022 · similarity 0.86)
- Learning Regional Attraction for Line Segment Detection (TPAMI 2019 · similarity 0.86)
- Improving Transformer Based Line Segment Detection with Matched Predicting and Re-ranking (AAAI 2025 · similarity 0.84)
- HomoMatcher: Achieving Dense Feature Matching with Semi-Dense Efficiency by Homography Estimation (AAAI 2025 · similarity 0.82)
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.
