Fadi Boutros

how many identities dose your training dataset has? you may consider a smaller batch size to fit your model in the GPU

Resnet 100 with 512 batch requires approximately 4 gpus each of 16GB. You need to set the batch size based on number of gpu and gpu memory

IResnet100 for both pretrained models

which database do you use for training, loss function, batch size etc...?

For casia dataset, you may need to train the model for at least 50 epochs with lr step of 20 30 and 40. The reported FR result in the paper...

You can use the pretrained model from [ElasticFace](https://github.com/fdbtrs/ElasticFace)

Hi, There is no restriction on the FIQ values range and they do not have to be within range of [0,1]. Not all approaches in the literature have FIQ value...

you need to +1 if you train the model. For prediction/testing, it is not needed