mtcnn-onnx-simple 0.4

MTCNN implementation using ONNXRuntime

ONNX MTCNN

Description

This project implements the MTCNN (Multi-task Cascaded Convolutional Networks) algorithm using ONNX (Open Neural Network Exchange) format. MTCNN is a popular face detection algorithm that consists of three stages: face detection, bounding box regression, and facial landmark localization. By using ONNX, we can leverage the benefits of interoperability and portability across different deep learning frameworks.

When you are going to use MTCNN to satisfy your face detection needs in your production environment, you can use this to get rid of the heavy Pytorch or Tensorflow dependency and instead opt for the lightweight ONNXruntime

Features

• Face detection using MTCNN algorithm
• Bounding box regression for accurate face localization
• Facial landmark localization for detailed facial analysis

Installation

1. Install the required dependencies: pip install MTCNN_ONNX

Usage

from mtcnn_onnx_simple import Align

align = Align()
image_path = "Image Path Here"
boxes, faces = align.get_aligned_faces(image_path)
# faces is an array of PIL Image Instances
for face in faces:
    face.show()

License

This project is licensed under the MIT License.