In this pandemic, we have lost numerous life because of negligence and not following guidelines. We are not focusing on the cause of such behaviour but rather than how to check such activities and prevent them from worsening as much as we can. With this kept in mind I've been working on our surveillance system, which would detect situation in real time where people are flaunting covid-19 rules and regulations.
What it does
As of now only two criteria are added into the system, which are 1. mask 2. social distance . If this rule are flaunted then an alert will be sent to the respective supervisor to act upon.
How we built it
We have built this app using pyqt5 and python mainly. For pedestrian detection we have used nanodet library and for face detection and mask detection we have used pytorch based implementation.
Challenges we ran into
The main and foremost necessary task was to be able to run our app in real time without any lag. For this we have leveraged the multi-threading pipeline of pyqt5 and perform all AI related calculation in a separate thread so that our main thread where our application event are running wont get affected.
Accomplishments that we're proud of
We are proud of
- that we have successfully integrated both pedestrian detection and mask detection in real time.
- displaying scatter graph of mask and pedestrian count with time
- giving user option to enter mobile and email to address/notify the provided supervisor whenever social distancing is not maintained 4.It will help organisation around the world to enforce covid-19 related rules.
What we learned
We have learned about real time model inferencing, Qt based application development .
What's next for covid-surviellance-desktop-app
Currently we can't have add more rules to our app, because of limitation of python's GLI lock which prevent us to use multi-threading to its fullest. Thus if provided an opportunity we will re-implement our whole application in cpp .And the notification tab is not working which still needs more work!