Decorrelated-Adversarial-Learning

Decorrelated-Adversarial-Learning copied to clipboard

## Resize transformer block was added ```python ImageFolderWithAgeGroup(pat=self.dataset['pat'], pos=self.dataset['pos'], \ cutoffs=age_cutoffs, root=self.dataset['train_root'], transform=transforms.Compose(\ [transforms.RandomHorizontalFlip(p=0.5), transforms.ToTensor(), \ transforms.Normalize((0.5,0.5,0.5), (0.5,0.5,0.5)), transforms.Resize((96, 112))])) ``` this was added to the valid as well

hasithz

Aboout Pretrained model

2

I use the pre-trained face recognition model to train ,the cano_cor is always 0. I think it may be related to the pre-training model. Would you like to provide your...

maryhh

about feature renorm

1

hi ~ i find you used norm feature in backbone , can you explain this op? And compared to un-norm feature ，is it usefully code: renorm(2,0,1e-5).mul(1e-5)

llf1234

Where is procedudre of the test, if I use the pretrained model on CACD-VS. And what is the data form of the test dataset? Thk.

xiaopanchen

Trained Model

1

Dear Can you please provide your trained model? I want to evaluate its performance on other datasets. I shall be grateful for your kind consideration, please. Thanks

saadshakeel12345

arcface or cosface?

1

Hello, Have you ever compared the effect of cosface and arcface？Which of them is better？ Thank you！

blackantywj

Some questions about data sets

1

Hello! How did you deal with the lack of age tags in VGGFace2? Can you please tell me what other data sets you used in addition to VGGface2 in the...

blackantywj

Is there any pre-processed dataset available? Thanks.

vateye