Research map / Detection and segmentation / CNN features
SOLD2: Self-Supervised Occlusion-Aware Line Description and Detection
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 |
|---|---|---|---|
| use joint self-supervised line detection and description | SOLD² · Wireframe and YorkUrban | line repeatability and localization | Helped |
| use SOLD² descriptors on LSD-detected lines instead of LBD | LSD lines · Wireframe | matching recall: 0.597 -> 0.889 | Helped |
| use SOLD² descriptors on LSD-detected lines instead of WLD | LSD lines · ETH3D | matching recall: 0.398 -> 0.525 | Helped |
| apply line NMS to remove overlapping SOLD² line segments (ablation) | SOLD² · ETH3D | matching recall: 0.538 -> 0.688 | Helped |
| apply line NMS to remove overlapping SOLD² line segments (ablation) | SOLD² · Wireframe | matching precision: 0.882 -> 0.777 | Hurt |
| replace ordered dynamic-programming sequence matching with averaged descriptors (ablation) | SOLD² line matching · Wireframe GT lines | matching accuracy: 0.978 -> 0.944 | Hurt |
| replace ordered dynamic-programming sequence matching with nearest-neighbor averaging (ablation) | SOLD² line matching · Wireframe GT lines with up to 50% occlusion | matching accuracy: 0.846 -> 0.803 | Hurt |
| use SIFT descriptors only at line endpoints (ablation) | SOLD² line matching · Wireframe GT lines | matching accuracy: 0.978 -> 0.532 | Hurt |
| sample line points by D2-Net saliency rather than regularly (ablation) | SOLD² line matching · Wireframe GT lines | matching accuracy: 0.978 -> 0.969 | Hurt |
| sample line points by ASLFeat saliency rather than regularly (ablation) | SOLD² line matching · Wireframe GT lines | matching accuracy: 0.978 -> 0.963 | Hurt |
| use 3 rather than 5 regularly sampled points per line (ablation) | SOLD² line matching · Wireframe GT lines with up to 50% occlusion | matching accuracy: 0.846 -> 0.813 | Hurt |
| use 10 rather than 5 regularly sampled points per line (ablation) | SOLD² line matching · Wireframe GT lines | matching accuracy: 0.978 -> 0.972 | Hurt |
Closest papers by meaning
- Learning Attraction Field Representation for Robust Line Segment Detection (CVPR 2019 · similarity 0.89)
- ScaleLSD: Scalable Deep Line Segment Detection Streamlined (CVPR 2025 · similarity 0.87)
- Improving Transformer Based Line Segment Detection with Matched Predicting and Re-ranking (AAAI 2025 · similarity 0.85)
- LINE: LLM-based Iterative Neuron Explanations for Vision Models (NeurIPS 2026 · similarity 0.84)
- MutualVPR: A Mutual Learning Framework for Resolving Supervision Inconsistencies via Adaptive Clustering (NeurIPS 2025 · similarity 0.83)
- SegMASt3R: Geometry Grounded Segment Matching (NeurIPS 2025 · similarity 0.83)
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.