Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain Adaptation for Semantic Segmentation With Maximum Squares Loss
Minghao Chen, Hongyang Xue, Deng Cai
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 |
|---|---|---|---|
| replace entropy minimization with maximum squares loss (ablation) | UDA segmentation · GTA5->Cityscapes | mIoU (%): 42.2 -> 44.3 | Improved |
| add image-wise weighting factor to maximum squares loss (ablation) | MaxSquare · GTA5->Cityscapes | mIoU (%): 44.3 -> 45.2 | Improved |
| add image-wise weighting factor to entropy minimization (ablation) | MinEnt · GTA5->Cityscapes | mIoU (%): 42.2 -> 43.5 | Improved |
| add multi-level self-produced guidance to MaxSquare (ablation) | MaxSquare · GTA5->Cityscapes | mIoU (%): 44.3 -> 45.2 | Improved |
| add multi-level guidance to MaxSquare+IW (ablation) | MaxSquare+IW · GTA5->Cityscapes | mIoU (%): 45.2 -> 46.4 | Improved |
| add image-wise weighting factor | MaxSquare · SYNTHIA->Cityscapes (16 classes) | mIoU (%): 39.3 -> 40.4 | Improved |
| add multi-level guidance | MaxSquare+IW · SYNTHIA->Cityscapes (16 classes) | mIoU (%): 40.4 -> 41.4 | Improved |
| use maximum squares loss instead of entropy minimization | ResNet-50 · Office-31 average | accuracy (%): 84.1 -> 85.6 | Improved |
| use maximum squares loss instead of entropy minimization | ResNet-50 · Office-31 A->W | accuracy (%): 89.0 -> 92.4 | Improved |
| use maximum squares loss instead of entropy minimization on hard samples (ablation) | fine-tuning · Office-31 A->W bottom set | accuracy | Improved |
| use maximum squares loss instead of entropy on easy samples (ablation) | fine-tuning · Office-31 A->W top set | accuracy | No clear effect |
| lower lambda_T from 0.1 to 0.02 (ablation) | MaxSquare · GTA5->Cityscapes | mIoU (%): 44.3 -> 43.0 | Hurt |
Closest papers by meaning
- ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation (CVPR 2019 · similarity 0.91)
- Unsupervised Domain Adaptation for Semantic Segmentation by Content Transfer (AAAI 2021 · similarity 0.91)
- Unsupervised Domain Adaptation for Semantic Segmentation via Class-Balanced Self-Training (ECCV 2018 · similarity 0.91)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.90)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.90)
- SemiDAViL: Semi-supervised Domain Adaptation with Vision-Language Guidance for Semantic Segmentation (CVPR 2025 · similarity 0.88)
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.
