Research map / Label-efficient and robust learning / Domain adaptation and generalization
Not All Areas Are Equal: Transfer Learning for Semantic Segmentation via Hierarchical Region Selection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| select synthetic pixels using hierarchical weighting networks | FCN on GTAV and Cityscapes · Cityscapes val | Mean IoU: 64.0% -> 68.1% | Helped |
| add hierarchical pixel-, region-, and image-level weighting (ablation) | FCN+W¹ on GTAV and Cityscapes · Cityscapes val | Mean IoU: 66.7% -> 67.6% | Helped |
| add hierarchical pixel-, region-, and image-level weighting (ablation) | FCN+W¹⁹ on GTAV and Cityscapes · Cityscapes val | Mean IoU: 66.9% -> 68.1% | Helped |
| use multi-channel rather than shared weighting maps | Hierarchical transfer learning on GTAV and Cityscapes · Cityscapes val | Mean IoU: 67.6% -> 68.1% | Helped |
| use hierarchical weighting with SYNTHIA source data | FCN on SYNTHIA and Cityscapes · Cityscapes val | Mean IoU: 65.3% -> 66.8% | Helped |
| add GTAV and SYNTHIA as synthetic sources | Hierarchical transfer learning · Cityscapes val | Mean IoU: 68.1% -> 68.8% | Helped |
| train jointly on synthetic and real images without source selection | FCN · Cityscapes val (GTAV + Cityscapes) | Mean IoU: 65.3% -> 64.6% | Hurt |
| add synthetic data through direct joint training | FCN · Cityscapes val (SYNTHIA + Cityscapes) | Mean IoU: 65.3% -> 62.9% | Hurt |
| use 1/2 rather than all real-world training images (ablation) | Ours with W¹⁹ using GTAV · Cityscapes val | Mean IoU: 68.1% -> 64.9% | Hurt |
| use 1/4 rather than all real-world training images (ablation) | Ours with W¹⁹ using GTAV · Cityscapes val | Mean IoU: 68.1% -> 57.7% | Hurt |
| use 1/8 rather than all real-world training images (ablation) | Ours with W¹⁹ using GTAV · Cityscapes val | Mean IoU: 68.1% -> 53.4% | Hurt |
Closest papers by meaning
- Learning Semantic Segmentation From Synthetic Data: A Geometrically Guided Input-Output Adaptation Approach (CVPR 2019 · similarity 0.91)
- Learning From Synthetic Data: Addressing Domain Shift for Semantic Segmentation (CVPR 2018 · similarity 0.90)
- Learning Texture Invariant Representation for Domain Adaptation of Semantic Segmentation (CVPR 2020 · similarity 0.90)
- Transferring and Regularizing Prediction for Semantic Segmentation (CVPR 2020 · similarity 0.90)
- ROAD: Reality Oriented Adaptation for Semantic Segmentation of Urban Scenes (CVPR 2018 · similarity 0.89)
- A Training-free Synthetic Data Selection Method for Semantic Segmentation (AAAI 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.