greenlab-library 0.1.1

HungLV package

[GreenLab] - Libraries

1. Description

This library is built for specific tasks

• Face Recognition Project
  • Face Detection
  • Extract Face Embedding
  • Search Face in database
• License Plate Recognition Project

Documentation:

2. Table of Contents

• Setup guide
• Usage
• Examples
• APIs
• References

3. Setup guide

3.1 System requirements

• Python>=3.6
• CUDA==10.0
• MXNet
• Tensorflow
• Keras

3.2 Installation

Make sure conda is installed.

Note that: If you are using GPU, you need to change the version of cuda in requirements.txt file.

mxnet-cu101 means the package is built with CUDA/cuDNN and the CUDA version is 10.1.

# create environment 
conda create --name face_recog_test python=3.6
conda activate face_recog_test

# install dependencies for cpu 
pip install -r requirements-cpu.txt
# or install dependencies for gpu 
pip install -r requirements-gpu.txt

# install face recognition library
pip install --upgrade green-face-recognition

4. Usage

Prepare the config file as yaml type. Take a look this example.

Show the list of models

from face_recognition import models
models.show_avai_models()

Load the models

retina_model = models.build_model('retina-r50',config_path)
arcface = models.build_model('arc-face',config_path)

Make the prediction

# get faces and landmarks
retina_model.detect_fast(img,img.shape,0.8,[1],do_flip=False)

5. Examples

Face Recognition - Getting Started

6. APIs

7. References

1. RetinaFaceModel

Maintainers

Author: Hung Le Viet

Last Update: July 14, 2020

Updates

• [Jul 9] Initial version
• [Jul 13] Make the models can run on GPU/CPU. Integrated rcnn library.