A implementation of Disentangled Makeup Transfer with Generative Adversarial Network.
-
Download MT (Makeup Transfer) dataset from here.
-
Put MT (Makeup Transfer) dataset to
.\data\RawData
. Your data path will like this:
.\data\RawData\images\makeup\*.png
.\data\RawData\images\non-makeup\*.png
.\data\RawData\segs\makeup\*.png
.\data\RawData\segs\non-makeup\*.png
-
run
python convert.py
-
Modify train.py and start training.
python train.py
-
run
python export.py
and you will get h5 model in.\Export
.
-
make sure you have run
python export.py
to get h5 model. -
Modify demo.py and run
python demo.py
, you will find the transfer result in.\Transfer
.
-
The test environment is
- Python 3.7
- tensorflow-gpu 2.0.0
- tensorflow-addons 0.7.1
- imgaug 0.4.0
-
This is still not a completed implementation, but almost 95% is the same as paper described.