Research map / Architectures and efficiency / RNNs and normalization
End-to-End Wireframe Parsing
Yichao Zhou, Haozhi Qi, Yi Ma
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 static line sampler (S+, S-) to random dynamic sampler (ablation) | L-CNN · ShanghaiTech ablation | sAP5: 43.7 -> 47.8 | Improved |
| add dynamic sampler candidates D+ and D- (and S+, S-) (ablation) | L-CNN · ShanghaiTech ablation | sAP5: 47.8 -> 58.9 | Improved |
| use dynamic sampler on predicted junctions instead of only static samples from ground truth (ablation) | L-CNN · ShanghaiTech ablation | sAP5: 38.5 -> 43.7 | Improved |
| add hard negative samples S- and D- (ablation) | L-CNN · ShanghaiTech ablation | sAP5: 57.4 -> 58.9 | Improved |
| add hard negative samples S- and D- (ablation) | L-CNN · ShanghaiTech ablation | sAP10: 61.4 -> 62.9 | Improved |
| replace conv+fc head with fc+fc head in line verification network (ablation) | L-CNN · ShanghaiTech ablation | sAP5: 55.7 -> 58.9 | Improved |
| replace conv+fc head with fc+fc head in line verification network (ablation) | L-CNN · ShanghaiTech ablation | sAP15: 61.7 -> 64.7 | Improved |
| train end-to-end wireframe parser (L-CNN) instead of Wireframe parser two-stage heuristic | L-CNN · ShanghaiTech test | sAP10: 5.1 -> 62.9 | Improved |
| L-CNN instead of AFM line detector | L-CNN · ShanghaiTech test | APH: 69.5 -> 83.0 | Improved |
| L-CNN instead of AFM line detector | L-CNN · ShanghaiTech test | FH: 77.2 -> 81.2 | Improved |
| L-CNN instead of AFM when tested on York Urban without finetuning | L-CNN · York Urban | sAP10: 3.5 -> 26.4 | Improved |
Closest papers by meaning
- Holistically-Attracted Wireframe Parsing: From Supervised to Self-Supervised Learning (TPAMI 2023 · similarity 0.92)
- NEAT: Distilling 3D Wireframes from Neural Attraction Fields (CVPR 2024 · similarity 0.85)
- Spot the Error: Non-autoregressive Graphic Layout Generation with Wireframe Locator (AAAI 2024 · similarity 0.82)
- COCOTree: A Dataset and Benchmark for Open Tree-Structured Visual Decomposition (NeurIPS 2026 · similarity 0.81)
- SceneSplat++: A Large Dataset and Comprehensive Benchmark for Language Gaussian Splatting (NeurIPS 2025 · similarity 0.80)
- CADTalk: An Algorithm and Benchmark for Semantic Commenting of CAD Programs (CVPR 2024 · similarity 0.80)
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.
