Research map / Detection and segmentation / Weakly supervised detection and segmentation
Revisiting the Necessity of Full Accuracy: Weakly Supervised Object-Level Offset Correction for Misaligned Building Labels
Junda Xu, Yanmeng Liu, Xiangqiang Zeng, Jinrong Wu, Ying Qu, Libao Zhang
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 |
|---|---|---|---|
| train segmentation on OMAF-corrected labels instead of raw offset labels | UNetFormer · Islahiye | mIoU (%): 35.8 -> 76.4 | Improved |
| train segmentation on OMAF-corrected labels instead of raw offset labels | Deeplabv3plus · Antakya | mIoU (%): 53.9 -> 66.0 | Improved |
| add self-alignment (edge and variance constraints) (ablation) | OMAF · Islahiye | mIoU (%): 57.12 -> 66.34 | Improved |
| add prior-based regularization (confidence weighting) (ablation) | OMAF · Islahiye | mIoU (%): 66.34 -> 67.50 | Improved |
| add instance-level offset regression network (ablation) | OMAF · Islahiye | mIoU (%): 67.50 -> 72.50 | Improved |
| add centroid positional feature to offset regression (ablation) | OMAF · Islahiye | mIoU (%): 72.50 -> 73.32 | Improved |
| use OMAF (<<1% manual labels) instead of semi-supervised DA with 20% labels | SegFormer · Islahiye | mIoU (%): 70.8 -> 77.3 | Improved |
| use OMAF instead of unsupervised domain adaptation | SegFormer · Antakya | mIoU (%): 48.6 -> 67.7 | Improved |
Closest papers by meaning
- Align and Segment: Unsupervised Learning for Building Segmentation From Misaligned Labels (ECCV 2026 · similarity 0.87)
- Learning to Extract Building Footprints From Off-Nadir Aerial Images (TPAMI 2022 · similarity 0.84)
- 3D Building Reconstruction from Monocular Remote Sensing Images with Multi-level Supervisions (CVPR 2024 · similarity 0.83)
- OmniCity: Omnipotent City Understanding With Multi-Level and Multi-View Images (CVPR 2023 · similarity 0.83)
- BuildingWorld: A Structured 3D Building Dataset for Urban Foundation Models (AAAI 2026 · similarity 0.82)
- Active Label Correction for Semantic Segmentation with Foundation Models (ICML 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.
