This repository implements a face parsing model using a modified BiSeNet (Bilateral Segmentation Network) architecture trained in PyTorch. The model performs pixel-level semantic segmentation on facial images, dividing them into distinct regions such as skin, eyes, mouth, hair, and other facial components. It is trained on the CelebAMask-HQ dataset and includes scripts for data preprocessing, distributed training, and evaluation. A companion project demonstrates using the parsing maps for face makeup transfer applications.