Skip to content

Latest commit

 

History

History
27 lines (22 loc) · 883 Bytes

README.md

File metadata and controls

27 lines (22 loc) · 883 Bytes
Raw

FMA-3D

A method for adding a mask on a non-masked face image. Given a real masked face image (a) and a non-masked face image (d), we synthesize a photo-realistic masked face image with the mask from (a) and the facial area from (d). image

Some Results by FMA-3D

image

Requirements

  • python >= 3.7.1
  • pytorch >= 1.1.0

Usage

  • Extract the landmarks. You can extract the 106 landmarks by our face sdk or any other methods.
  • Add a mask on a face image. You can refer to add_mask_one.py as an example.
python add_mask_one.py

Speed Up

Some advice:

  • Write the whole process by multi-processing.
  • Write the function of render in face_masker.py by c++.

Reference

This project is mainly inspired by PRNet.