Research map / Label-efficient and robust learning / Domain adaptation and generalization
Constructing Self-Motivated Pyramid Curriculums for Cross-Domain Semantic Segmentation: A Non-Adversarial Approach
Qing Lian, Fengmao Lv, Lixin Duan, Boqing Gong
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 pixel-square middle layers to top + bottom pyramid (ablation) | PyCDA (ResNet-101) · GTAV to Cityscapes | mIoU (%): 46.2 -> 47.4 | Improved |
| add pixel pseudo labels (bottom) to top-layer label distribution (ablation) | PyCDA (ResNet-101) · GTAV to Cityscapes | mIoU (%): 42.0 -> 46.2 | Improved |
| add image-level label distributions (top) to pixel pseudo labels (ablation) | PyCDA (ResNet-101) · GTAV to Cityscapes | mIoU (%): 40.6 -> 46.2 | Improved |
| combine CDA with ST at pixel level instead of either alone (ablation) | PyCDA (VGG-16) · GTAV to Cityscapes | mIoU (%): 28.1 -> 34.9 | Improved |
| replace superpixels with pixel squares (ablation) | PyCDA (ResNet-101) · GTAV to Cityscapes | mIoU (%): 46.3 -> 46.3 | No clear effect |
| use pixel squares in the middle layer in addition to pixel labels instead of top + squares only (ablation) | PyCDA (ResNet-101) · GTAV to Cityscapes | mIoU (%): 46.3 -> 47.4 | Improved |
| add pixel-square middle layers on SYNTHIA (ablation) | PyCDA (ResNet-101) · SYNTHIA to Cityscapes | mIoU (%): 44.8 -> 46.7 | Improved |
| add pixel-square middle layers on SYNTHIA (ablation) | PyCDA (VGG-16) · SYNTHIA to Cityscapes | mIoU (%): 35.1 -> 35.9 | Improved |
| apply PyCDA domain adaptation to source-only model | FCN8s VGG-16 · GTAV to Cityscapes | mIoU (%): 24.3 -> 37.2 | Improved |
| apply PyCDA domain adaptation to source-only model | PSP-Net ResNet-101 · SYNTHIA to Cityscapes | mIoU (%): 33.0 -> 46.7 | Improved |
| use 16x16 square middle layer instead of 4x4 and 8x8 squares (ablation) | PyCDA (ResNet-101) · GTAV to Cityscapes | mIoU (%): 47.4 -> 47.5 | No clear effect |
| use very large 128x128 squares in the pyramid (ablation) | PyCDA (ResNet-101) · GTAV to Cityscapes | mIoU (%): 47.5 -> 47.0 | Hurt |
Closest papers by meaning
- A Curriculum-Style Self-Training Approach for Source-Free Semantic Segmentation (TPAMI 2024 · similarity 0.92)
- Unsupervised Intra-Domain Adaptation for Semantic Segmentation Through Self-Supervision (CVPR 2020 · similarity 0.91)
- Domain Adaptive Semantic Segmentation With Self-Supervised Depth Estimation (ICCV 2021 · similarity 0.91)
- A Curriculum Domain Adaptation Approach to the Semantic Segmentation of Urban Scenes (TPAMI 2019 · similarity 0.91)
- Self-Supervised Augmentation Consistency for Adapting Semantic Segmentation (CVPR 2021 · similarity 0.91)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.89)
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.
