snowfallingplum / fpmt

FPMT: Fast and Precise High-Resolution Makeup Transfer via Frequency Decomposition

Requirements

A suitable conda environment named FPMT can be created and activated with:

conda env create -f environment.yaml
conda activate FPMT

Pretrained Models

The pre trained FPMT with L=3 model has been provided in folder weights. Due to the parameter of FPMT being only 0.95M, its pth file size is approximately 3.6M.

Run the examples

In the current directory, run the following command

python eval.py

The corresponding generated results have been placed in folder examples_results\test_pair

How to run your own data

1. Face alignment. Please refer to the BeautyGAN or PSGAN code to crop the image according to the face landmarks.

2. Prepare face parsing. In our experiment, face parsing is generated by https://github.com/zllrunning/face-parsing.PyTorch.

3. Put face images in the ./examples/images/makeup and ./examples/images/non-makeup. Put the results of face parsing in the ./examples/seg1/makeup and /examples/seg1/non-makeup.

4. Run python eval.py.