Siamese network and minimal data quantity

Hello, can anyone help? I’m trying to create siamese network for faces verification, i used AT&t faces dataset, it’s different 40 classes per face, one class inslude 10 small pictures for the particular face. I tried siamese network withh CNN and FC layers, but i guess there’s not enough data for a training, cause i got overfitting for 10th epoch already. Can anyone give an admit about data quantity?

1 Like

Is this somehow related to any OpenAI models?

Hi!
You should head on over to the Huggingface community.

That’s likely a better place for good answers regarding this type of question.