zxhuang97
commented
5 years ago Gram loss is not used in this model. In the training phase, it's just an autoencoder. The purpose of the underlined function is to extract the required style features.
Open yongxiongwei opened 5 years ago
zxhuang97
commented
5 years ago Gram loss is not used in this model. In the training phase, it's just an autoencoder. The purpose of the underlined function is to extract the required style features.
you not use compute_style_features function ,this function compute gram matrix.