[NeurIPS 2024] Learning Frequency-Adapted Vision Foundation Model for Domain Generalized Semantic Segmentation

This is the official implementation of our work entitled as Learning Frequency-Adapted Vision Foundation Model for Domain Generalized Semantic Segmentation, which has been accepted by NeurIPS 2024.

Environment Configuration

Please refer to the requirements.txt file in this project.

Training on the Source Domain

An command example to train the model when using CityScapes as the source domain is:

python tools/train.py configs/my/citys_rein_dinov2_mask2former_512x512_bs1x4.py --work-dir exps/exp0322

Please remember to change the file folder to your own.

Inference on the Unseen Target Domains

An command example to infer the model when using CityScapes as the source domain is:

python tools/test.py configs/my/citys_rein_dinov2_mask2former_512x512_bs1x4.py exps/exp0429/iter_40000.pth --backbone checkpoints/dinov2_converted.pth

Please remember to change the file folder to your own, and also to specify the file folder of the target domain in the script.

Citation

If you find this work is useful for your task please cite our work as follows:

@inproceedings{bi2024fada,
  title={Learning Frequency-Adapted Vision Foundation Model for Domain Generalized Semantic Segmentation},
  author={Bi, Qi and Yi, Jingjun and Zheng, Hao and Zhan, Haolan and Huang, Yawen and Ji, Wei and Li, Yuexiang and Zheng, Yefeng},
  booktitle={Advances in Neural Information Processing Systems (NeurIPS)},
  volume={37},
  year={2024}
}

Acknowledgement

Our implementation is primarily based on the following repositories, with significant influence from Rein. Thanks for their authors.

• MMSegmentation
• Rein