informaticacba / face-mask-detection-using-yolov3

This is to detect face mask and non masked faces apart in realtime

YOLOv3 based Mask and non-mask identifier

About

It is simple implementation of YOLOv3 model for detecting masked perople apart from unmasked ones in real time. Thid sydtem can be useful for entrance in this pandemic.

Useful Links

Do follow following links for dataset, model, and trained weight file.

Dataset From Kaggle

Trained Weight file

YOLO darknet from AlexeyAB

Requirements

• Python version 3.8

• Python libraries listed in requirements.txt

    pip install -r requirements.txt
  

• Weight file. ( If you do not want to train on your own )

Training

Follow training steps as one by one from Training_YOLOv3.ipynb file. ( Use of Google colab suggested )

Inference

First you should download Weight file and keep it in the same folder.

• To test on image execute test_on_photo.py file.

    python test_on_photo.py
  

• To test on image execute test_on_video.py file.

    python test_on_video.py
  

Results

Outputs

Images

Video

result.mp4

Training Curve

Currently I have trained upto 2400 iterations and got descent result. this model can be trained even more. Try training upto 6000 iterations or even more.

Output parameters

mAP : 76%

fps on CPU : 1.7fps

Training Time : 2hrs 13 min

---