Research map / Detection and segmentation / Open-set and few-shot detection
Learning to Extract Building Footprints From Off-Nadir Aerial Images
Jinwang Wang, Lingxuan Meng, Weijia Li, Wen Tao Yang, Lei Yu, Gui-Song Xia
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 offset head (ablation) | Mask R-CNN · BONAI test | footprint F1-score: 58.12 -> 61.78 | Improved |
| add offset head (ablation) | PANet · BONAI test | footprint F1-score: 58.06 -> 62.15 | Improved |
| add offset head (ablation) | HRNetv2-W32 · BONAI test | footprint F1-score: 60.81 -> 63.16 | Improved |
| add offset head (ablation) | CM R-CNN · BONAI test | footprint F1-score: 60.94 -> 63.73 | Improved |
| add FOA module (ablation) | LOFT · BONAI test | footprint F1-score: 61.78 -> 64.31 | Improved |
| add FOA module (ablation) | LOFT · BONAI test | roof F1-score: 67.17 -> 67.25 | No clear effect |
| add FOA module (ablation) | LOFT · BONAI test | average end-point error (pixel): 5.26 -> 4.94 | Improved |
| use four rotation angles instead of one in FOA (ablation) | LOFT w/ FOA · BONAI test | footprint F1-score: 61.78 -> 64.31 | Improved |
| share FC layers (not Conv) across FOA branches (ablation) | LOFT w/ FOA · BONAI test | footprint F1-score: 63.84 -> 64.31 | Improved |
| share Conv layers across FOA branches too (ablation) | LOFT w/ FOA · BONAI test | footprint F1-score: 64.31 -> 62.94 | Hurt |
| replace FOA by image-level rotation augmentation at same 24 epochs (ablation) | LOFT · BONAI test | footprint F1-score: 64.31 -> 59.88 | Hurt |
| train image-level rotation augmentation longer (ablation) | LOFT with IRA · BONAI test | footprint F1-score: 59.88 -> 64.43 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PANet — PANet: Few-Shot Image Semantic Segmentation With Prototype Alignment (ICCV 2019) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning Building Extraction in Aerial Scenes with Convolutional Networks (TPAMI 2017 · similarity 0.87)
- 3D Building Reconstruction From Monocular Remote Sensing Images (ICCV 2021 · similarity 0.84)
- Revisiting the Necessity of Full Accuracy: Weakly Supervised Object-Level Offset Correction for Misaligned Building Labels (CVPR 2026 · similarity 0.84)
- Structured Outdoor Architecture Reconstruction by Exploration and Classification (ICCV 2021 · similarity 0.84)
- 3D Building Reconstruction from Monocular Remote Sensing Images with Multi-level Supervisions (CVPR 2024 · similarity 0.83)
- Aerial Lifting: Neural Urban Semantic and Building Instance Lifting from Aerial Imagery (CVPR 2024 · similarity 0.81)
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.
