Retinaface

A simple package of face detection

This package is built on top of the Retinaface

• More about Retinaface

• Installation

• Usage

• [How to do a python package ](#How to do a python package)

More about Retinaface

Retinaface is the State-of-the-art for Face Detection on WIDER Face. There are two versions of retinaface: MobileNet Backend and Resnet Backend. The model using MobileNet as backbone has only 1.7M, the other model with Resnet backbone has ~30m. Here is the performance on the FDDB dataset:

Installation

$ pip install pytorch-detection

Usage

from retinaface import RetinafaceDetector
import cv2 as cv


### Mobinet backbone 
detector  = RetinafaceDetector(net='mnet').detect_faces
img  = cv.imread('./imgs/DSC_8221.jpg')
bounding_boxes, landmarks = detector(img)
print(bounding_boxes)

### Resnet backbone 
detector  = RetinafaceDetector(net='mnet').detect_faces
img  = cv.imread('./imgs/DSC_8221.jpg')
bounding_boxes, landmarks = detector(img)
print(bounding_boxes)

Result

How to make a python package

To create this project locally, create the following file structure:

retinaface
    ├── retinaface
    │      ├── __init__.py
    ├── LICENSE
    ├── README.md
    ├── setup.py
    ├── setup.py
    ├── requirements.txt

Creating setup.py

setup.py is the build script for setuptools. It tells setuptools about your package

import setuptools

setuptools.setup(
    name="Pytorch-detection", # Replace with your own username
    version="0.0.1",
    author="Hoang Phuong",
    author_email="hphuongdhsp@gmail.com",
    license='MIT',
    description = "A simple example package for face detection",
    long_description = long_description,
    long_description_content_type="text/markdown",
    url="https://github.com/hphuongdhsp/retinaface",
    packages=setuptools.find_packages(),
    classifiers=[
        "Programming Language :: Python :: 3",
        "License :: OSI Approved :: MIT License",
        "Operating System :: OS Independent",
    ],
    keywords='face detection, retinaface',
    install_requires=open('requirements.txt').readlines(),
    python_requires='>=3.6',
)

Creating README.md

Creating a LICENSE

Generating distribution archives

Install setuptools and wheel

$ pip install --user --upgrade setuptools wheel

make distribution files:

python3 setup.py sdist bdist_wheel

That command genetate two files in the dist directory: