Inspiration

The current COVID-19 crisis and the need for people to wear masks in public spots to ensure the safety of both themselves and others.

What it does

An application to monitor CCTV footage and alert(via a bounding box) when people are not wearing face mask to CCTV operators in a locally hosted webpage that displays the processed stream.

How I built it

Use a Custom Vision exported TensorFlow model that can then be run locally in the edge on the host and data is sent to the user in the browser when the URL is opened in real time.

Why

  • This project abstracts away the need for users to build a system to capture and process video streams and display it on a dashboard
  • Instead the user only needs to train a production ready model on customvision.ai
  • User needs to only provide the RTSP URLs, username and password
  • Users can train their own object detection model using their own dataset on customvision.ai and export the TensorFlow model that can then be used with the project

Challenges I ran into

  • Architecting the multithreading for processing of multiple video streams

What I learned

  • Integrating CV into python
  • Using Custom Vision exported models in python
  • Using flask and flask-socketio

Built With

Share this project:

Updates