torch-fidelity VS trajectopy-core

The Fréchet Inception Distance (FID) is a widespread metric to assess the quality of the distribution of a image generative model (GAN, Stable Diffusion, etc.). The metric is not trivial to implement as one needs to compute the trace of the square root of a matrix. In all PyTorch repositories I have seen that implement the FID (https://github.com/mseitzer/pytorch-fid, https://github.com/GaParmar/clean-fid, https://github.com/toshas/torch-fidelity, ...), the authors rely on SciPy's sqrtm to compute the square root of the matrix, which is unstable and slow.

Try torch-fidelity : https://github.com/toshas/torch-fidelity

You can, but two ingredients are required: a rich pretrained feature extractor for your data to replace InceptionV3 pretrained on ImageNet, and https://github.com/toshas/torch-fidelity, which can be used with custom feature extractors (also supports FID and KID).

The first, called trajectopy, stands as a full-fledged application featuring a PyQt6-based graphical user interface (GUI). This GUI-driven platform simplifies trajectory-related tasks and offers an intuitive user experience. For those desiring a more in-depth approach, there is trajectopy-core. This backend implementation without any PyQt6 dependencies provides essential functionality e.g. for computing absolute trajectory error (ATE) and relative pose error (RPE).