- A tensorflow Implement for BeautyGAN:Instance-level facial makeup transfer with deep generative adversarial networks.
- Based on cycleGAN tf Implement
- Baidu Drive for vgg16.npy: 83dt
- face parsing tools: dlib'68 landmarks model
- Another MakupTransfer Model based on BeautyGAN and AdaIn
- face parsing is not precise -> remove face mask/ add total variation loss
- in order to speed up the training process, we fix the (content,style) pair. remove this constraint should improve the model's performance on validation data and test data.
- we simply apply the face parser in opencv, the original paper use different face parsing tool.
