Face-Mask-Detector

A face mask detector based on STM32F103ZET6 and Yolov4.

Interface

Compile and Load

You should install arm-none-eabi-gcc to compile this project.

cd stm32
make update

Build Environment

1. Create virtual environment:

   conda create -n Face_Mask_Detector python=3.8
   conda activate Face_Mask_Detector
   pip install -r requirements.txt
   

2. Install PyTorch, refer to the blog for details.

Train

1. Download face mask dataset from kaggle and unzip it.

2. Download pre-trained CSPDarknet53.pth model from Google Drive.

3. Modify the value of root in train.py, please ensure that the directory structure of the root folder is as follows:

   root
   ├───Annotations
   ├───ImageSets
   │   ├───Layout
   │   ├───Main
   │   └───Segmentation
   ├───JPEGImages
   ├───SegmentationClass
   └───SegmentationObject
   

4. start training:

   conda activate Face_Mask_Detector
   python train.py
   

Evaluation

one model

1. Modify the value of root and model_path in eval.py.

2. Calculate mAP:

   conda activate Face_Mask_Detector
   python eval.py
   

multi models

1. Modify the value of root and model_dir in evals.py.

2. Calculate and plot mAP:

   conda activate Face_Mask_Detector
   python evals.py
   

mAP curve

Detection

1. Modify the model_path and image_path in demo.py.

2. Display detection results:

   conda activate Face_Mask_Detector
   python demo.py
   

License

Face-Mask-Detector is licensed under GPLv3.

Copyright © 2021 by zhiyiYo.