mini-face

Installation with pip

Requirements: python >= 3.12, numpy >= 1.21.

1. Install the package with a following command:

   python3 -m pip install mini-face

2. Download model data files (model and AU_predictors) from this repository or official OpenFace release. You will also need files from OpenFace models.

Example

To estimate gaze on a single image, simply run:

from pathlib import Path

import cv2
import numpy as np
from mini_face import gaze, PredictionMode


if __name__ == "__main__":
    extractor = gaze.Extractor(
        mode=PredictionMode.IMAGE,         # Specify prediction mode
        focal_length=(500.0, 500.0),       # Specify camera parameters
        optical_center=(860.0, 540.0),     # required for proper 3D estimation
        models_directory=Path("./model"),  # Pass a directory to search for OpenFace weights in
    )

    image = cv2.imread("test_image.jpg")
    result = extractor.predict(
        image,                             # Pass the image to analyze
        np.array([0, 0, 1080, 1920])       # along with a region to search for face in
    )

    print(f"{result = }")