Research map / Label-efficient and robust learning / Domain adaptation and generalization
GA-DAN: Geometry-Aware Domain Adaptation Network for Scene Text Detection and Recognition
Fangneng Zhan, Chuhui Xue, Shijian Lu
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 |
|---|---|---|---|
| adapt IC13 training images with GA-DAN | EAST · ICDAR2015 | F-score: 53.3 -> 64.4 | Improved |
| adapt IC13 training images with GA-DAN | EAST · MSRA-TD500 | F-score: 46.8 -> 58.5 | Improved |
| generate 10 adapted images per source image instead of 1 (multi-modal spatial learning) (ablation) | EAST · ICDAR2015 | F-score: 64.4 -> 69.4 | Improved |
| generate 10 adapted images per source image instead of 1 (multi-modal spatial learning) (ablation) | EAST · MSRA-TD500 | F-score: 58.5 -> 62.1 | Improved |
| add 10x adapted IC13 images to target training data | EAST · ICDAR2015 | F-score: 80.0 -> 83.5 | Improved |
| add 10x adapted IC13 images to target training data | EAST · MSRA-TD500 | F-score: 70.3 -> 75.5 | Improved |
| use GA-DAN instead of CycleGAN for adaptation | EAST · ICDAR2015 | F-score: 57.2 -> 64.4 | Improved |
| use GA-DAN instead of ST-GAN + CycleGAN concatenation | EAST · ICDAR2015 | F-score: 60.8 -> 64.4 | Improved |
| add disentangled cycle-consistency loss (vs normal cycle loss) (ablation) | GA-DAN + MORAN · COMB-CUTE | accuracy: 32.6 -> 38.2 | Improved |
| add disentangled cycle-consistency loss (vs normal cycle loss) (ablation) | GA-DAN + MORAN · COMB-SVTP | accuracy: 44.9 -> 47.1 | Improved |
| add appearance adaptation to spatial-only output (ablation) | GA-DAN + MORAN · COMB-CUTE | accuracy: 36.1 -> 38.2 | Improved |
| generate 10 adapted images per source image instead of 1 (ablation) | GA-DAN + MORAN · COMB-CUTE | accuracy: 38.2 -> 43.1 | Improved |
Closest papers by meaning
- Geometry Normalization Networks for Accurate Scene Text Detection (ICCV 2019 · similarity 0.87)
- Geometry-Aware Scene Text Detection With Instance Transformation Network (CVPR 2018 · similarity 0.87)
- Joint Geometrical and Statistical Alignment for Visual Domain Adaptation (CVPR 2017 · similarity 0.85)
- Dual Bipartite Graph Learning: A General Approach for Domain Adaptive Object Detection (ICCV 2021 · similarity 0.85)
- Beyond Batch Learning: Global Awareness Enhanced Domain Adaptation (TPAMI 2025 · similarity 0.85)
- Geometry-Aware Network for Domain Adaptive Semantic Segmentation (AAAI 2023 · 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.
