search

williamyang1991 / GP-UNIT

[CVPR 2022] Unsupervised Image-to-Image Translation with Generative Prior
Other
192 stars 15 forks source link

How to calculate the FID and LPIPS #3

Closed zjwulbx closed 1 year ago

zjwulbx commented 2 years ago

Dear authors: I want to ask about the evaluation FID and LPIPS. When we calculate the FID and LPIPS, what dataset do we use? and under what circumstances(such as some details as how many pictures do we use? Using training dataset or testing dataset? What are the sources of real_images and fake_images, e.g. cat2dog)? Can you tell me more details about evaluation? Looking forward to you early reply!

williamyang1991 commented 2 years ago

We use and modify this code for evaluation https://github.com/clovaai/stargan-v2/blob/875b70a150609e8a678ed8482562e7074cdce7e5/metrics/eval.py

Fake images are generated from the testing set of the dataset. For cat2dog, there are 500 testing images (https://github.com/clovaai/stargan-v2/blob/master/README.md#animal-faces-hq-dataset-afhq).

For FID, the real images are the training images in the dataset.

zjwulbx commented 2 years ago

Thank you for your answer