The pandemic may end but the lifestyle it has introduced is here to stay. The hygiene and sanitary habits induced by the onset of the pandemic are soon to become the norm. Moreover, a strict lockdown can’t be imposed for prolonged spells of time. Eventually, markets would have to be opened and workplaces would be allowed to function under strict regulations in order to sustain human life and well-being. This pandemic poses a global crisis and we must dedicate our brainpower to combating it. Now in the absence of any vaccine or cure, the only strategy is to diligently practice social distancing. Through our project, we wish to encourage such norms.
What it does
The project observes people in workplaces or other places susceptible to crowds and monitors how well people are following the social distancing norms. Our project also gives real-time analytics to identify potential hotspots as soon as possible. This could also be used to compare Social Distancing patterns in different parts of the world.
How we built it
We used OpenCV's built in DNN algorithm along with YoloV3 Darknet with the COCO dataset to detect pedestrians. And constructed our own image distance algorithms.
Challenges we ran into
- Measuring distances on an image
- Detecting herds of pedestrians walking together.
Accomplishments that we're proud of
Finally coming up with an acceptable image distance measurement algorithm.
What we learned
How to operate complex object detection models such as YoloV3 along with OpenCV's built in Neural Network. We also learned how to plot animated graphs using matplotlib
What's next for Safe Crowds
Incorporating AI powered facial recognition algorithms to pull up the details from a database of the people identified in the video and sending alerts if necessary. Refining distance measurement in images using stereo camera or dual camera algorithms. Incorporating cough-detection machine learning models and body temperature detection using IR technology to detect individuals who should isolate themselves.