Face recognition models like VGG Face are designed to have a classification head on top and then trained to classify face images, but after they are trained the classification head can be removed and facial embeddings can be extracted and used for face recognition and face verification tasks on previously unseen data.

My question is, rather than design a classification model, can a regression model be designed to predict the facial embeddings vector directly ?

I tried something like that but the model always converges to the mean value and doesn’t learn anything, Although this can be due to poor hyper parameter choices or unsuitable model architecture.

I am not specifically asking about facial embeddings, any embeddings will do for my question.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.