liorshk / facenet_pytorch Goto Github PK

Hello,
would you please tell why do it like this?

Why I cannot access the dataset link here?

PermissionError: [Errno 13] Permission denied: '/media/lior'

Hi @liorshk

Thanks for sharing your code. It seems clean and well-written, however, I had problem having it converge.

I trained it on filtered version of MsCeleb with 5 Million images and 79K identities. Your hyper-parameters seems to be identical with the Tensorflow implementation davidsandberg/facenet and I also tried different ones but I never got more than 65% accuracy on LFW.

I think it's mostly because of the way that triplet selection has been implemented. The paper suggests having batches of 1800 images from a certain number of identities (40-45), rather than choosing it completely randomly. I tried this but only with 180 images at most, yet still it didn't converge.

Do you have any idea that can help me? If you had any success training the model, could you please share your weights too?

Thanks,

when I train triplet loss, I find bug in here:
out_selected_a = Variable(torch.from_numpy(out_a.cpu().data.numpy()[hard_triplets]).cuda())
out_selected_p = Variable(torch.from_numpy(out_p.cpu().data.numpy()[hard_triplets]).cuda())
out_selected_n = Variable(torch.from_numpy(out_n.cpu().data.numpy()[hard_triplets]).cuda())

it should be out_a[hard_triplets]. Otherwise, grad won't be caculated.

Why do you create pairwise function instead of pytorch's nn.PairwiseDistance?

I changed pairwise function with nn.PairwiseDistance but evaluating is not working. (Too slow!)
Do you have any idea about this?

out_selected_a = Variable(torch.from_numpy(out_a.cpu().data.numpy()[hard_triplets]).cuda())
out_selected_p = Variable(torch.from_numpy(out_p.cpu().data.numpy()[hard_triplets]).cuda())
out_selected_n = Variable(torch.from_numpy(out_n.cpu().data.numpy()[hard_triplets]).cuda())

will lead to

RuntimeError: element 0 of tensors does not require grad and does not have a grad_fn

why not use:

out_selected_a = out_a[hard_triplets]
out_selected_p = out_p[hard_triplets]
out_selected_n = out_n[hard_triplets]

Hello, I try to run your code. However, i cannot find the download link for, would you mind sharing me this link. In addition, i want to check is the link of lfw-align-mtcnn in your train_center.py http://vis-www.cs.umass.edu/lfw/lfw-deepfunneled.tgz? Thank you so much.

Dear @liorshk,
Thank you for your code. Do you have any pre-trained models? Would you please, release the weights of your pre-trained models?

Hello, author, can you tell me the running environment of this code?Examples include the Python version and the PyTorch version.