Inspiration
The rampaging outbreak of COVID-19, commonly known as the corona virus
What it does
Our product is our own machine learning model which uses facial recognition and hand gestures to detect a person who is coughing severely in a public space through a CCTV footage. The current prototype mainly works in airports, as cross-country travel is the main cause of the global spread of the virus. Based on the intensity of the persons cough, the facial recognition accesses the individuals travel history through the airport’s database and if the person if found to be travelling to highly infested areas he will be dealt with by the airport authorities.
How I built it
We trained our own custom datasets of hand gestures through the open source deep Neural Network, darknet, by manually plotting coordinates and training thousands of pictures. Then through facial landmark detection we combined our machine learning model and could detect intensive coughing in real time. Once recognized, our facial recognition database provided a history of the identified user’s travel history and based on risk levels, conducted emergency checkups for extra precaution.
Challenges I ran into
This was our first experience with deep Neural Networks and to build a highly accurate machine learning model without any thermal camera, infrared sensors etc. but only using a simple laptop webcam.
Accomplishments that I'm proud of
This is by far the most unique project we’ve worked on and are proud of creating the world’s first ever visual cough detection model
What's next for Coro-nah
Our product is unique because there is no such algorithm which detects coughing without the use of a thermal or an infrared camera. Instead of terabytes of CCTV data from public spaces going to waste we have found a Big Data solution most recent dangers on the world population. It product is an open source code which can be regulated by the governments as this is a privacy issue and we do not intent to breach any privacy laws in any legislations around the world
Log in or sign up for Devpost to join the conversation.