Details missing regarding the FID computation

[koutilya-pnvr]

Thanks for the amazing work. However there are some details missing around the computation of FID in tables 1 and 2 of your paper.

Can you give more information for the following:

1. How many generated samples you computed the FID for?
2. How much of the real dataset you used during the above FID computation on generated images?

Thanks so much for your time and help.

[yumingj]

Hi @koutilya-pnvr

Sorry for the late reply. We use 1149 generated samples when computing FID. The real dataset also contains 1149 images. The images are from the test dataset.

Thanks! Hope the above information addresses your concerns.

[soon-yau]

Can you provide your generated test samples for download?