minchul

/

cvlface_adaface_ir101_webface12m

like 4

Hi,
I was testing this pretrain and I realized that the input processing is different from 2 source:

1. HuggingFace is using RGB channel
2. Github is using BGR and in the README note that the input is BGR channel.

So, what is the true type of input that the pretrain use ?
Thanks,