Inspiration
We were inspired by the world’s slow recovery from the COVID-19 pandemic. Lockdown is being relaxed but life will not be back to normal so soon. As business & community open up, it becomes more important than ever that people stay safe and avoid transmitting the virus. We thought that there needs to be a system that minimizes the spread of COVID-19 and also to increase awareness among the citizens. COVID-19 has been shown to remain airborne on various surfaces for extended periods of time. We were thinking that how would workplaces run efficiently and take necessary safety precautions as the lockdown cools down in many countries though the covid 19 virus is still an issue. And this is where we came up with the idea of c19 watcher
What it does
C19 Watcher is an application that is suitable for Post Pandemic work Environment.Our Goal is to achieve a safe work Environment where all the social norms such as Social Distancing, Wearing a Mask, and checking body temperature is done through minimal human interaction. For this we have created various Machine Learning and Deep Learning models which can be implemented on any camera module and all of this is integrated into a fast serverless application known as C19 Watcher. The C19 Watcher application is capable of identifying the person without a mask and sending his/her contact details to the authority which can then take necessary actions also the application has an added feature by which the authority can use the application for calling the person or pinging him/her on Whatsapp directly. As for the Social Distancing practise whenever someone violates it an image will be captured by our Model and sent to security for further intervention.
How We Built It!
C-19 Watcher is an app backed by a flutter and firebase architecture.The frontend was created using flutter and Dart. We have deployed our Machine Learning and Deep Learning model on Google Cloud Platform,which is capable of capturing live updates from the camera and sending the notification to administration. For the authentication and storing user essential details we have used the firebase and firestore database respectively. We have deployed three models that are social distancing detector, face mask detector and face recognition. In the social distancing model we have used Facebook's detectron2 model for object detection and localization, this gives us a bounding box around all the people in the frame and then we used the scipys spatial model for calculating the distance between all the people out of which the minimum distance is taken and if that is less than the threshold an alert is sent to the database. The second and third models are integrated together so the model returns the person's name who is not wearing a mask along with his photo, for face mask detection we have used mobilenet's model for object classification and using opencv we are creating a bounding box around the face. And for face recognition we used adam geitgey's face recognition module.
Challenges We ran into
The main problem we faced was in the integration of the application and the Deep Learning Models.Firstly the Face recognition model data was not reaching the database when it reached the database and the database wasn't loading the captured image. Our plan was to send the image of the people violating social distancing to the database but the number of frames being sent were 150 in a 5 seconds video.So now we had to remove the almost similar frames.So we thought of comparing the last frame sent, to the previous ones, that way we can remove the unnecessary frames.
Accomplishments that We're proud of
We are extremely proud of C19 watchers overall as it has the potential to make a difference in everyone’s worlds. It also has enabled workplaces throughout to work safely and with assurance. In particular we are proud of the aesthetic user interface of our app and our model will help the people to stay safe and also help the organization to run at a faster rate with etiquette prevention from COVID-19.Lets make going to workplace safe and fun again!
What We learned
During the Hackathon, we learned to work as a team. Connecting online was a bit of a challenge but it was an experience worth remembering. Since we had so many sponsors we were able to learn firebase APIs, integrating ML models with REST APIs and integrating with firebase storage.
What's next for C19 Watcher
For Our application we have thought of introducing an important feature in respect to the corona virus that is temperature. We are thinking of measuring the person’s temperature using thermal imaging which can help prevent spreading of virus to a great extent.
Log in or sign up for Devpost to join the conversation.