Tiny RetinaFace

Tiny RetinaFace is a fast and simplified implementation of biubug6/Pytorch_Retinaface inference module

Installation

pip install tiny-retinaface

Usage

from retinaface import read_image, detect_faces

# Read image as RGB ndarray
image = read_image("images/test.jpg")

# Detect faces as list of Face objects
faces = detect_faces(image)

Face object

Face(
    score=0.9997380375862122,
    x=142,
    y=322,
    width=60,
    height=72,
    landmarks=[
        Point(x=160, y=348),
        Point(x=187, y=345),
        Point(x=177, y=362),
        Point(x=165, y=376),
        Point(x=188, y=374)
    ]
)

Tiny RetinaFace Results on World Largest Selfie

from retinaface import read_image_url, detect_faces, draw_faces

url = "https://github.com/yakhyo/retinaface-pytorch/blob/main/assets/large_selfi.jpg?raw=true"

image = read_image_url(url)
faces = detect_faces(image, score_threshold=0.1, nms_threshold=0.1)

draw_faces(image, faces, filename="images/results.jpg")

Acknowledgements

• Original RetinaFace implementation: deepinsight/insightface
• Pytorch RetinaFace implementation: biubug6/Pytorch_Retinaface
• RetinaFace pretrained onnx weights: yakhyo/retinaface-pytorch

References

@inproceedings{Deng2020CVPR,
title = {RetinaFace: Single-Shot Multi-Level Face Localisation in the Wild},
author = {Deng, Jiankang and Guo, Jia and Ververas, Evangelos and Kotsia, Irene and Zafeiriou, Stefanos},
booktitle = {CVPR},
year = {2020}
}

License

This project is licensed under the MIT License. See the LICENSE file for details.