facial-analysis

Face Analysis (ONNX models)

This repository contains functionalities for face detection, age and gender classification, face recognition, and facial landmark detection. It supports inference from an image or webcam/video sources.

Features

• Face Detection: Utilizes Sample and Computation Redistribution for Efficient Face Detection (SCRFD) for efficient and accurate face detection.
• Gender & Age Classification: Provides discrete age predictions and binary gender classification (male/female).
• Face Recognition: Employs ArcFace: Additive Angular Margin Loss for Deep Face Recognition for robust face recognition.
• Facial Landmark Detection
• Real-Time Inference: Supports both webcam and video file input for real-time processing.

Installation

1. Clone the repository:

git clone https://github.com/yakyo/facial-analysis.git
cd facial-analysis

1. Install the required dependencies:

pip install -r requirements.txt

1. Download weight files:

   a) Download weights from the following links:

   Model	Weights	Size
   ArcFace	w600k_r50.onnx	166 MB
   SCRFD	det_2.5g.onnx	3.14 MB
   SCRFD	det_500m.onnx	2.41 MB
   SCRFD(default)	det_10g.onnx	16.1 MB
   GenderAge	genderage.onnx	1.26 MB

   b) Run the command below to download weights to the weights directory (Linux):

   sh download.sh
   

Usage

python main.py --source assets/in_video.mp4
                        assets/in_image.jpg
                        0 # for webcam

main.py arguments:

usage: main.py [-h] [--detection-weights DETECTION_WEIGHTS] [--attribute-weights ATTRIBUTE_WEIGHTS] [--source SOURCE] [--output OUTPUT]

Run face detection on an image or video

options:
  -h, --help            show this help message and exit
  --detection-weights DETECTION_WEIGHTS
                        Path to the detection model weights file
  --attribute-weights ATTRIBUTE_WEIGHTS
                        Path to the attribute model weights file
  --source SOURCE       Path to the input image or video file or camera index (0, 1, ...)
  --output OUTPUT       Path to save the output image or video

Reference

1. https://github.com/deepinsight/insightface