prepare.md

Preparation Instructions

Clone this repository and prepare the dataset and weights through the following steps:

a. Prepare model weights for face detection.

Download the weights of dlib face detector of 68 landmarks here. Unzip it and move it to the directory ./faceutils/dlibutils.

Download the weights of BiSeNet (PyTorch implementation) for face parsing here. Rename it as resnet.pth and move it to the directory ./faceutils/mask.

b. Prepare Makeup Transfer (MT) dataset.

Download raw data of the MT Dataset here and unzip it into sub directory ./data.

Run the following command to preprocess data:

python training/preprocess.py

Your data directory should look like:

data
└── MT-Dataset
    ├── images
    │   ├── makeup
    │   └── non-makeup
    ├── segs
    │   ├── makeup
    │   └── non-makeup
    ├── lms
    │   ├── makeup
    │   └── non-makeup
    ├── makeup.txt
    ├── non-makeup.txt
    └── ...

c. Download weights of trained EleGANt.

The weights of our trained model can be download here. Put it under the directory ./ckpts.