search

mysee1989 / TCAE

Self-supervised Representation Learning from Videos for Facial Action Unit Detection
162 stars 27 forks source link

About the dataset #3

Closed wangyr915 closed 4 years ago

wangyr915 commented 5 years ago

As mentioned in your paper, you preprocessed the training dataset "VoxCeleb" with face detection, alignment. So if we want to retrain the model, we have to do the same process on the Voxceleb dataset? Thank you!

mysee1989 commented 5 years ago

Yes, you should do the same proccess, i.e., face detection, face alignment & crop.