My dataset contains single pairs of images of the same identity. That is, (Selfie, Identity Card with face on it). From this, I would like to calculate the performance of my facial recognition system (FMR, FNMR). However from the sources I read, it is expected that I have multiple pairs/images of the same person to get a list of genuine scores. Which I do not have
I would be training my face rec (ML) model, so I understand that I would have to create a few fake images of the same identity (using GANs) to produce a multi sample dataset. However, I would like to establish a benchmark FMR and FNMR first, using an existing pretrained face rec model.
Any ideas how an issue can be resolved?