This repository contains the code to compute the Fréchet Video Distance (FVD) as described in https://arxiv.org/abs/1812.01717.
FVD is implemented in tensorflow and requires tensorflow
and tensorflow-hub
to be available. An example is available here.
Given a tensor of real videos, and a tensor of generated videos that both have dimensionality [NUMBER_OF_VIDEOS, VIDEO_LENGTH, FRAME_WIDTH, FRAME_HEIGHT, 3]
with values in 0-255, FVD is computed as
from frechet_video_distance import frechet_video_distance as fvd
result = fvd.calculate_fvd(
fvd.create_id3_embedding(fvd.preprocess(real_videos, (224, 224))),
fvd.create_id3_embedding(fvd.preprocess(generated_videos, (224, 224))))
If you are using FVD please cite us!
@article{unterthiner2018towards,
title={Towards Accurate Generative Models of Video: A New Metric \& Challenges},
author={Unterthiner, Thomas and van Steenkiste, Sjoerd and Kurach, Karol and Marinier, Raphael and Michalski, Marcin and Gelly, Sylvain},
journal={arXiv preprint arXiv:1812.01717},
year={2018}
}