face-parsing.PyTorch

Our fork scripts:

• EPE_seg_infer.py - takes input directory with images, predicts segmentation masks for them and saves masks to the given output directory
• EPE_inference_with_detector_* - scripts for testing detectors, takes input directory with images and saves detection results to given output_directory
• extra_video2frames.py - extract frames from given video and save them to "frames" directory in our dataset structure
• extra_copyframes_afterwards.py - copy frames from one dataset to another, both datasets must have our dataset structure

Contents

• Training
• Demo
• References

Training

1. Prepare training data: -- download CelebAMask-HQ dataset

   -- change file path in the prepropess_data.py and run

python prepropess_data.py

2. Train the model using CelebAMask-HQ dataset: Just run the train script:

    $ CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2 train.py

If you do not wish to train the model, you can download our pre-trained model and save it in res/cp.

Demo

1. Evaluate the trained model using:

# evaluate using GPU
python test.py

Face makeup using parsing maps

face-makeup.PyTorch

	Hair	Lip
Original Input
Color

References

• BiSeNet