SkylabSORT

Risk of asymptomatic transmission of highly infectious disease such as COVID-19 is increased due to social interaction among unknown carriers. This problem is amplified in densely populated areas such as cities. Contact tracing is an extremely hard task to carry out especially if the population is being infected at an exponential rate. However, cities have densely-connected networks of CCTVs monitoring citizens on a day-to-day basis. By creating a tracking and re-identification model, the government is able to automatically track the whereabouts of an infected citizen before he/she was confirmed to have the virus. This will allow the relevant agencies to swiftly trace and isolate necessary targets to slow the spread of the virus. If deployed globally, people can be tracked even if they have traveled and the number of imported-cases will be drastically reduced. Punishments can also be exacted on those who have found to have flouted the law. 

We are tackling the Challenge on all fronts, helping healthcare workers fight the virus by giving them precious information on the whereabouts of people who are confirmed to have the virus. This problem of asymptomatic community spread is so prevalent in Singapore that the government employs part-time "Safe Distancing Ambassadors" to separate people in public. People who have found to have broken the law by going out unnecessarily, not wearing a mask in public, or flouting their "Stay Home Notice" have been fined or jailed (caught more than 3 times). Utilizing CCTV data and machine learning is an excellent tool to protect the public as well as to speed up the work of contact tracing. More manpower and man-hours can be spent on fighting the virus instead. In conclusion, we are able to detect and "follow" the spread of the virus through movement of people. Then, swift measures can be taken to isolate and treat the infected to slow the spread of the virus.

The solution is implemented using a deep learning model on datasets collected from live CCTVS around the country. When a person with the disease is identified, the model will be deployed to identify others that the sick person has interacted with. This can prevent the spread of a pandemic by identifying those with potential carriers of the disease and isolating them before the spread intensifies. Through this solution, contact tracing can be deployed in a scalable way with a low manpower overhead.

The target population SkyLab is working to improve would be the team of contact tracers, working to identify who may have been at risk of contracting a highly contagious disease, an example of which would be Covid-19. One way we, as students. have been doing is to watch videos on how contact tracers have been tracking down the close contacts of those who have contracted the disease. We understand that using data analysis and comparisons can be be a mentally taxing and laborious process for them. Hence, it is hoped that our conceptual prototype would be able to make their search process less menial, so they are able to focus their energy on contacting close contacts. The computer vision solution we are proposing would allow the contact tracers to not only know who would have been at risk of contracting, it also allows them to ascertain their identity, resulting in quicker processes of isolation. Ultimately, it is hoped that our solution can bring about a lower community spread of the current, and future pandemics.

With current data of infections and CCTV footage, we can train re-identification models to accurately predict how people interact and transmit the virus to others, preparing for similar outbreaks. This autonomous solution will enable human manpower to be used more efficiently, like identifying family members or providing aid to vulnerable communities.

We will track those who could be infected and alert officials, to detect and test them swiftly. Millions can be traced and identified without manpower involved and requires no expensive hardware or training to use, allowing for swift global response and poorer nations to do this with little friction.

Person Re-Identification through camera tracking is a technology proven to work and have been developed heavily in the past decade. Here are a few papers that showcase a wide variety of methods 

4

We are Engineering Science students from Ngee Ann Polytechnic and our final year project is quite relevant.

We do not have a business model