ELDA

Official implementation of ELDA: Using Edges to Have an Edge on Semantic Segmentation Based UDA

Ting-Hsuan Liao, Huang-Ru Liao, Shan-Ya Yang, Jie-En Yao, Li-Yuan Tsao, Hsu-Shen Liu, Bo-Wun Cheng, Chen-Hao Chao, Chia-Che Chang, Yi-Chen Lo, Chun-Yi Lee
BMVC 2022 | arXiv |

Installation

pip3 install -r requirements.txt

Data Preparation

Download Cityscapes, GTA5 and SYNTHIA-RAND-CITYSCAPES. And change the corresponding data path in ./data/__init__.py .

Training

Model will be saved under the ./saved directory.

1) GTA5 -> Cityscapes

python3 trainUDA_gta.py --config ./configs/configUDA_gta2city_edge.json --name ELDA_gta

1) SYNTHIA -> Cityscapes

python3 trainUDA_synthia.py --config ./configs/configUDA_syn2city_edge.json --name ELDA_synthia

Evaluate

1) GTA5 -> Cityscapes

python3 evaluateUDA.py --full-resolution -m deeplabv2_gta --model-path ./checkpoint/gta5_edge_exp/checkpoint.pth

Change [--model-path] to the corresponding checkpoint path.

1) SYNTHIA -> Cityscapes

python3 evaluateUDA.py --full-resolution -m deeplabv2_synthia --model-path ./saved/synthia_egde_exp/checkpoint.pth

Change [--model-path] to the corresponding checkpoint path.

Acknowledgement

• CorDA is used as our codebase.

Citation

Please cite our work if you find it useful.

@inproceedings{eldauda,
  title={ELDA: Using Edges to Have an Edge on Semantic Segmentation Based UDA},
  author={Ting-Hsuan Liao, Huang-Ru Liao, Shan-Ya Yang, Jie-En Yao, Li-Yuan Tsao, Hsu-Shen Liu, Bo-Wun Cheng, Chen-Hao Chao, Chia-Che Chang, Yi-Chen Lo and  Chun-Yi Lee},
  booktitle={BMVC},
  year={2022}
}