Research map / Detection and segmentation / Weakly supervised detection and segmentation
Align and Segment: Unsupervised Learning for Building Segmentation From Misaligned Labels
Venkanna Babu Guthula, Oswin Krause, Dimitri Gominski, Hui Zhang, Johan Mottelson, Ankit Kariryaa, Nico Lang, Christian Igel
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 consistency loss L_Con (ablation) | AnS (L_CE) · Las Vegas, random misalignment | IoU_seg: 0.71 -> 0.74 | Improved |
| Add consistency loss L_Con (ablation) | AnS (L_CE) · Las Vegas, systematic misalignment | IoU_seg: 0.39 -> 0.41 | Improved |
| Add geometric augmentation alone (ablation) | AnS (L_CE) · Las Vegas, random misalignment | IoU_seg: 0.71 -> 0.66 | Hurt |
| Add geometric augmentation alone (ablation) | AnS (L_CE) · Las Vegas, systematic misalignment | IoU_seg: 0.39 -> 0.72 | Improved |
| Add both L_Con and augmentation (ablation) | AnS (L_CE) · Las Vegas, systematic misalignment | IoU_align: 0.41 -> 0.88 | Improved |
| Add both L_Con and augmentation (ablation) | AnS (L_CE) · Las Vegas, random misalignment | IoU_seg: 0.71 -> 0.79 | Improved |
| Train TNet and SNet end-to-end instead of two-stage with frozen pretrained TNet (ablation) | AnS · ReBO test | IoU_seg: 0.38 -> 0.62 | Improved |
| Freeze pretrained SNet encoder instead of fine-tuning (ablation) | AnS SNet (ConvNeXt DINOv3) · systematic misalignment (Las Vegas) | IoU_seg: 0.41 -> 0.69 | Improved |
| Use AnS instead of unsupervised baselines (MA, AC, SC) | AnS · Las Vegas, systematic misalignment | IoU_align: MA 0.47, AC 0.46 -> AnS 0.88 | Improved |
| Use AnS instead of supervised MapRepair | AnS · Paris, random misalignment | IoU_align: MR 0.57 -> AnS 0.67 (random); MR 0.68 -> AnS 0.65 (systematic) | Mixed |
Closest papers by meaning
- Universal Semi-Supervised Semantic Segmentation (ICCV 2019 · similarity 0.87)
- Undoing the Damage of Label Shift for Cross-Domain Semantic Segmentation (CVPR 2022 · similarity 0.87)
- Improving Semi-Supervised Semantic Segmentation with Sliced-Wasserstein Feature Alignment and Uniformity (CVPR 2025 · similarity 0.86)
- Label-Efficient Few-Shot Semantic Segmentation with Unsupervised Meta-Training (AAAI 2024 · similarity 0.86)
- Active Label Correction for Semantic Segmentation with Foundation Models (ICML 2024 · similarity 0.86)
- AllSpark: Reborn Labeled Features from Unlabeled in Transformer for Semi-Supervised Semantic Segmentation (CVPR 2024 · similarity 0.86)
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.
