Face Detection For Python using ONNX

This package implements parts of Google®'s MediaPipe models in pure Python (with a little help from Numpy and PIL) without Protobuf graphs and with minimal dependencies (just ONNX and Pillow).

Credits

This is a fork of a repo by patlevin who did all the heavy lifting. Credits to patlevin for the majority of this code.
Original repository: https://github.com/patlevin/face-detection-tflite

Models and Examples

The package provides the following models:

• Face Detection

• Face Landmark Detection

• Iris Landmark Detection

• Iris recoloring example

Motivation

The package doesn't use the graph approach implemented by MediaPipe and is therefore not as flexible. It is, however, somewhat easier to use and understand and more accessible to recreational programming and experimenting with the pretrained ML models than the rather complex MediaPipe framework.

Here's how face detection works and an image like shown above can be produced:

from fdlite import FaceDetection, FaceDetectionModel
from fdlite.render import Colors, detections_to_render_data, render_to_image 
from PIL import Image

image = Image.open('group.jpg')
detect_faces = FaceDetection(model_type=FaceDetectionModel.BACK_CAMERA)
faces = detect_faces(image)
if not len(faces):
    print('no faces detected :(')
else:
    render_data = detections_to_render_data(faces, bounds_color=Colors.GREEN)
    render_to_image(render_data, image).show()

While this example isn't that much simpler than the MediaPipe equivalent, some models (e.g. iris detection) aren't available in the Python API.

Note that the package ships with five models:

• FaceDetectionModel.FRONT_CAMERA - a smaller model optimised for selfies and close-up portraits; this is the default model used
• FaceDetectionModel.BACK_CAMERA - a larger model suitable for group images and wider shots with smaller faces
• FaceDetectionModel.FULL - a model best suited for mid range images, i.e. faces are within 5 metres from the camera

If you don't know if you need FULL of BACK_CAMERA, use the BACK_CAMERA as it is most versatile.

Installation

The latest version be installed via:

pip install git+https://github.com/seppe-intelliprove/face-detection-onnx

The package can be also installed from source by navigating to the folder containing pyproject.toml and running

pip install .

Development, Building and publishing

This project uses python-poetry for dependency management. It can also be used to build and publish this package.

install required dependencies

poetry install

build a wheel

poetry build