search

CyberAgentAILab / layout-dm

LayoutDM: Discrete Diffusion Model for Controllable Layout Generation [Inoue+, CVPR2023]
https://cyberagentailab.github.io/layout-dm
Apache License 2.0
211 stars 23 forks source link

FID computation #24

Closed modric197 closed 1 year ago

modric197 commented 1 year ago

May I ask how many samples do we need when computing FID? Since FID score is very sensitive to the number of samples. Thank you!

naoto0804 commented 1 year ago

In general, the larger, the better. However, the number of samples in the test subset is limited. In our case, thought I'm not sure if I remember correctly, it was about 4k for Rico and about 10k for PubLayNet.