AI Based COVID 19 Solution Compliance

AI Based Mask Detection, Social Distancing Detection, Thermal Camera Fever Detection, Red Eye Detection

Comment

The problem AI-Based COVID 19 Solution Compliance solves We have built an AI-Based Product which serves as a solution for Covid-19 Pandemic Situation. It consists of AI-based screening of people who are having symptoms of Covid19 such as Fever and Conjunctivitis. The second module is the face mask detection system, and the third one is social distancing alert raising system. This is a scalable software product which could withstand strong and weak light and could give accurate results even in extreme temperature and humidity.

✔This is our AI Product to check and screen out the person who are not following the rules or carrying infection symptoms. ✔Contrary to others, this system is scalable and can detect multiple streams in optimal time. ✔Can be deployed on edge and cloud both along with battery backup(12 hrs max.) ✔The USP of this project is that it can detect multiple people (around 100) in one frame. ✔This has great use in surveillance and security purpose to find out who all are following the rules

Symptoms Detection ✔This snippet of software aims at checking the temperature of the human face and detect fever along with people who are having conjunctivitis. ✔Conjunctivitis(Red Eye) and Fever, both are considered as symptoms of Covid19 as per the guidelines of the World Health Organization ✔This has great use in surveillance and security purpose to find out who all are potentially infected by COVID 19 ✔ This can be used in Schools, Hospitals, Offices, Railway Stations and Airports

Mask Detection This is our AI Product to check if the person is wearing a mask or not. ✔Red: Mask, Pink: No_Mask ✔This has great use in surveillance and security purpose to find out who all are following the rules

Social Distancing Alert System WHO has advised maintaining Social Distancing at Workplaces ✔This system raises a trigger alert whenever someone breaks social distancing in the monitored area ✔We have measured the distance from the bottom midpoint, rather than the exact mid-point of the detection box. ✔Can be used in Isolation

Challenges we ran into Rendering of the frame was very slow without optimisation Detecting multiple people in one frame was a challenge Heavy neural network causes heating of edge device, we overcame this using tensor-RT optimisation Colab Notebooks have a limit of 12 hrs GPU, difficulty in training large models Collecting data of Indian people with beards for detection.

Built With

Share this project:

Updates