The problem The Detector solves

The ramifications COVID-19 has had on human-life have been on an unprecedented scale. This pandemic has single-handedly managed to disrupt entire industries and businesses. Many unfortunate souls with families to support, lost their livelihood and many more lost their lives to this pandemic. In view of all these losses, governments of many nations, on advice from doctors and scientists, came up with several remedies to curb the spread of this virus. Among other solutions, the more common ones include - social-distancing and wearing masks. In spite of these measures taken by the government, some people continue to be adamant and disregard these rules. Also to keep a track of it and to ensure that everyone abides by it is a challenging task. People have been spotted not wearing masks and some even continue to visit public places in large groups.

Our solution

In light of all these problems our team has engineered an elegant solution; one that could help deal with problems related to social distancing. It can be helpful for the public to maintain these guidelines and prevent themselves from being infected.

Mobile application - Expects a photo as input from the public. The various algorithms in place analyze them and identify the number of people in the image who are:

1. Not social-distancing.
2. Not wearing masks.

Web application - Organizations and educational institutes can use this application in their installed CCTV cameras to ensure that people are following the rules strictly. The data would then be specific to these organizations only.

Both frameworks have a centralized backend system for storing and dynamically updating the database. This ensures real-time security and awareness. We truly believe that our solution can help workplaces and make public places accessible and safe again!

Challenges we ran into

• Issue: We needed to add 2 Machine Learning models in a mobile application and create real-time graphs in the website and mobile application and for that to happen smoothly we required a centralized database system for both the platforms which was difficult.

• Solution: We made a centralized backend in Flask and connected the Mongo DB database for saving our data. Mongo DB is deployed on Cloud which is easier to be integrated into different front-end frameworks.

• Issue: Detecting human faces in crowded images and to calculate the distance between the detected humans was a difficult task.

• Solution: We used Haarcascade classifier with OpenCV to detect faces from images for the detection of masks on them. YOLO v3 object detection model was used to detect humans and we used the OpenCV library to find the distance between detected people.