This is an implementation of VGG-16 Wasserstein Distortion in PyTorch.
Wasserstein Distortion was introduced in:
Y. Qiu, A. B. Wagner, J. Ballé, L. Theis: "Wasserstein Distortion: Unifying Fidelity and Realism," 2024 58th Ann. Conf. on Information Sciences and Systems (CISS), 2024. https://arxiv.org/abs/2310.03629
For a description of this implementation, refer to:
J. Ballé, L. Versari, E. Dupont, H. Kim, M. Bauer: "Good, Cheap, and Fast: Overfitted Image Compression with Wasserstein Distortion," 2025 IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), 2025. https://arxiv.org/abs/2412.00505
Please cite these papers if you use this code for scientific work.
A more comprehensive implementation in JAX can be found in CoDeX.
Yueyu Hu (@huzi96) and Jona Ballé (@jonaballe).
pip install git+https://github.com/balle-lab/wasserstein-distortion.git
Check wasserstein_distorion_test_example.py.