Research map / Detection and segmentation / Weakly supervised detection and segmentation
Weakly Supervised Segmentation of Small Buildings With Point Labels
Jae-Hun Lee, ChanYoung Kim, Sanghoon Sull
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 size-separated loss with small object mask (whole image, no point sampling) over size-weighted U-net | RU+wsize · CrowdAI Mapping Challenge | All F-1: 0.8053 -> 0.8295 | Improved |
| add uncertainty point sampling to Ours-I (ablation) | Ours-I · CrowdAI Mapping Challenge | small-object recall: 0.4551 -> 0.6068 | Improved |
| add uncertainty point sampling to Ours-I (ablation) | Ours-I · CrowdAI Mapping Challenge | small-object precision: 0.3132 -> 0.2702 | Hurt |
| add uncertainty sampling to Ours-I (ablation) | Ours-I · CrowdAI Mapping Challenge | All F-1: 0.8295 -> 0.8319 | Improved |
| add memory bank and false positive sampling to Ours-P-U (ablation) | Ours-P-U · CrowdAI Mapping Challenge | small-object precision: 0.2702 -> 0.3607 | Improved |
| add memory bank and false positive sampling to Ours-P-U (ablation) | Ours-P-U · CrowdAI Mapping Challenge | All F-1: 0.8319 -> 0.8494 | Improved |
| use proposed method instead of ScribbleNet | ScribbleNet · CrowdAI Mapping Challenge | All F-1: 0.4895 -> 0.8494 | Improved |
| use proposed Ours-P-USB instead of RU+wsize | RU+wsize · WHU building | All F-1: 0.8208 -> 0.8490 | Improved |
| use proposed Ours-P-USB instead of RU+wsize | RU+wsize · Massachusetts buildings | All F-1: 0.5413 -> 0.5669 | Improved |
| use proposed Ours-P-USB instead of RU+wsize, large-object precision on densely packed buildings | RU+wsize · Massachusetts buildings | large-object precision: 0.3932 -> 0.4956 | Improved |
| use proposed Ours-P-USB instead of RU+wsize | RU+wsize · CrowdAI Mapping Challenge | small-object precision: 0.2593 -> 0.3607 | Improved |
| use a smaller point-label radius r (ablation) | Ours-P-USB · CrowdAI Mapping Challenge | performance | Hurt |
Closest papers by meaning
- Weakly- and Semi-Supervised Panoptic Segmentation (ECCV 2018 · similarity 0.87)
- Weakly Supervised Segmentation on Outdoor 4D Point Clouds With Temporal Matching and Spatial Graph Propagation (CVPR 2022 · similarity 0.87)
- Semi-Supervised Semantic Image Segmentation With Self-Correcting Networks (CVPR 2020 · similarity 0.87)
- Weakly Supervised Semantic Point Cloud Segmentation: Towards 10x Fewer Labels (CVPR 2020 · similarity 0.87)
- Aerial Lifting: Neural Urban Semantic and Building Instance Lifting from Aerial Imagery (CVPR 2024 · similarity 0.85)
- WISH: Weakly Supervised Instance Segmentation using Heterogeneous Labels (CVPR 2025 · similarity 0.85)
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.
