Covid-19-Face-Mask-and-Social-Distancing-Detection

This project is used to detect face mask and social distancing violations from input image, video and webcam feeds.

Models and Techniques

Pretrained Yolov3 is used to detect people

Pretrained SSD is used to detect faces

Trained MobileNetV2 is used as face mask classifier

Euclidean distance is used to calculate social distancing violations

Getting Started

Prerequisites

• face-detection
• flask
• imutils
• keras
• matplotlib
• opencv-python
• pandas
• scikit-learn
• tensorflow
• opencv-python

Quickstart (Demo)

Download the pretrained Yolov3 weights using this link and save it to the yolo-coco/ directory

Create a virtual environment and install the required dependencies using the command pip install -r requirements.txt

Face mask detection on images

Test the face mask classifier using the command python detect_face_mask_image.py --image input/image/path

The output image is stored as output_fm.png

You can also use the flask web app using the command python app.py and the application runs in the address http://0.0.0.0:12000/

Social distancing detection on images

Test the social distancing detection on images using the command python detect_social_distance_image.py --image input/image/path --distance [default=100.0]

Experiment --distance value for different images. The output is stored as output_sd.jpg

Face mask and social distancing detection on videos/webcam

Use the command python video.py --video input/video/path --distance [default=100.0] --frames [default=20] to test on video files

Experiment --distance value for different video files and --frames to skip frames. The result frames are stored in result_frames/ directory

Use the command python webcam.py to test using a webcam device

Docker for Face Mask Detection on Images

Use the Docker image to run the face mask detector microservice

Pull the docker image using the command docker pull rakeshraj97/project1:0.0.1

Run the docker using the command docker run -p 12000:12000 rakeshraj97/project1:0.0.1

Ensure working of the microservice using the command curl http://0.0.0.0:12000/ or open the link http://0.0.0.0:12000/ in a web browser to use the web application

Train Face Mask Detector

Dataset description

The dataset used to train face mask detector can be downloaded using this link

This is a balanced dataset containing faces with and without masks with a mean height of 283.68 and mean width of 278.77

Train

Use the command python train_mask_detector.py --dataset input/dataset/path to train the face mask classifier

Output of the Trained Model

Face Mask detection

Social Distancing detection

References

*https://www.pyimagesearch.com/2020/06/01/opencv-social-distancing-detector/

*https://www.pyimagesearch.com/2020/05/04/covid-19-face-mask-detector-with-opencv-keras-tensorflow-and-deep-learning/