Face APP Beautify

A face beautify / makeup transfer demo and inference workflow.

Installation

Install the required dependencies:

pip install numpy opencv-python pillow tqdm onnxruntime

Model

Download the models from the following link: models face_models.pth

Place the downloaded model files (including detector, parsing, and transfer ONNX weights) under the project model/ directory so that paths such as model/HP_FACE_DETECTOR.onnx, model/faceparsing.onnx, and model/appface_transfer_female/ / model/appface_transfer_male/ resolve correctly.

Usage

1. Put test images in inputs/.
2. Run the main entry script:

appface.py

Example:

python appface.py -t 0 -d 1 -f "[398, 233, 715, 645]"

3. Results are written under outputs/.

Main Parameters

• -t / --type: processing gender preset, int, 0 or 1, default 0
  • 0: female style set (model/appface_transfer_female/)
  • 1: male style set (model/appface_transfer_male/)
• -d / --detail: sharpness / detail strength, float, default 1
  • 1: default model behavior
  • 1.1–1.5: extra sharpening layered on top
• -f / --face_rect: optional face bounding box, string parsed as a four-number list [x1, y1, x2, y2]
  • If omitted, all detected faces above the confidence threshold are processed.

Requirements

Runtime Python packages are listed in the Installation section above (numpy, opencv-python, pillow, tqdm, onnxruntime).