The problem your project solves
The lack of solutions for contagion risk monitorization without privacy violation.
The solution you bring to the table
A plug-and-play software, to be deployed using any video recording device, combining people behavior with external variables in order to provide an accurate contagion risk evaluation of the area under surveillance.
What you have done during the weekend
During the weekend we implement a full functional software that can evaluate the COVID-19 contagion risk. We split the team on 3 main sub-teams: Health Support, Data Base/Layout and Data Science/Computer Vision. Each sub-team was autonomous, what led us to obtain faster results.
Health Support: Performed a scientific research work regarding how several variables (Temperature, Variations in Temperature, Relative Humidity, Social Distance, Ventilation, Sunlight hours and Air Pollution) affect the spreading of the SARS-CoV-2. All the information was matched in order to set three numerical ranges for each variable. Each of these ranges was defined into “high risk” at which the infection risk is high, “intermediate” at which the risk of infection is intermediate, and “optimum” stage at which probability of infection is lower.
Data Base/Layout: Creation of a database with all the variables that come from our data processing application and display them in a browser interface. Server installation and administration. Generation of a heat map using the object location inputs. Managing information threshold and developing a user-friendly interface for our service.
Data science/Computer vision: Developing an object detection API that attributes coordinates to the people detected through surveillance cameras with the YoloV3 model. Taking the image and transforming it into a bird eye perspective with an homography technique so that each person is in a realistic and scalar position. Then an analysis of the distance between them is possible. Parallel to this the information was combined with external variables and was treated in order to generate the contagion risk value.
The solution’s impact to the crisis
When the lockdown is over, there will be a need to secure public spaces. In that note, we propose a system that can provide a smart monitorization of human flows and public interactions. Our solution will allow a smooth transition between the lockdown and the return to our everyday life. Also, it can be easily apply and massively distributed, being specially designed to be implement in adversarial conditions like this pandemic crisis.
The necessities in order to continue the project
The Camlion software core is done, the next steps would be to find a partner that would be interested in deploying it. A list of future possible add-ons would require a full time team with a health specialists and data science/computer vision analysts. A extensively data collection of this specific virus will be necessary to improve our risk of contagion calculations. Working with adapted image data-set will be a priority in order to train our own neural network and improve the detection results. Hardware would be necessary so that the project becomes less time-consuming and so we can run more complex experiments.
The value of your solution(s) after the crisis
Camlion software can suffer some modifications in order to be applied to a variety of services and utilities.
Our particular system can be used to identify human flows in order to optimize the spatial organization of venues. It can also be used to gather data about human behavior in a large scale keeping the privacy.
The plug-and-play nature of the system, makes it easier to be massively applied. We can also set up the system to identify the contamination risk of a normal cold, and like that prevent the risk of contamination with other virus. Also, we could integrate physical features so we could identify gender, age or health problems.