How are the ear masks generated ?
sahilsid opened this issue · 5 comments
sahilsid commented
The roi-tanh warping model for face segmentation trained on the helen dataset does not give the segmenation for the ear region. How is this achieved ?
sicxu commented
We just use the roi-tanh warping method training on a dataset with annotations for ears.
sahilsid commented
@sicxu Could you tell me what is the dataset that you used for training roi-tanh ?
sicxu commented
CelebA-HQ
brunovollmer commented
Is the training code for the roi-tanh method published somewhere?
sicxu commented
Sorry for the inconvenience, there is no offical implementation of roi-tanh method, @cravisjan97 mentioned an implementation here #6 (comment). For more details about the method, please refer to Lin et al. 's paper.