SFEW about dan HOT 4 CLOSED

Thanks for your interest !
We only make public for the codes of RAF-DB and AffectNet datasets. As for SFEW 2.0, the images need to be crop manually, and the different face detection model may lead in various result with our code.

from dan.

May I know if you use the same transformations as in affectnet or raf-db? i.e:

from dan.

Not really, It's the transfom function, you can make a referance:

data_transforms = transforms.Compose([
        transforms.Resize((224, 224)),
        transforms.RandomHorizontalFlip(),
        transforms.RandomApply([
            transforms.RandomAffine(20, scale=(0.8, 1), translate=(0.2, 0.2))
            ], p=0.5),
        transforms.ToTensor(),
        transforms.Normalize(mean=[0.485, 0.456, 0.406],
                                 std=[0.229, 0.224, 0.225]),
        ])

from dan.

Thank you. May I know how do you crop and which face detection model you used in order to get your high performance?

from dan.