Detecting Potentially Infectious Disease with Deep Learning

A system that detects and tracks potential infectious disease carriers through surveillance cameras using deep learning techniques.

Prof. Dr. Ir. Sim Kok Swee

The standard of operating procedure (SOP) to reduce the risk of COVID-19 infection rate includes the process of acquiring the body temperature of the visitors in every entrance of the building. This process requires extra manpower and often causes congestion of the visitor queue.  Hence, we propose a system that combines the surveillance camera with a thermal sensor to track the body temperature of visitors. The facial information and body temperature are uploaded to the online database through cloud services.

Another issue faced during this pandemic is to accurately track the person who has closed contact with the COVID-19 carrier. Hence, we propose a system which employs several deep learning techniques to recognize the face of the infected person (subject) from the recorded video, track the paths where the subject travelled before, and determine the crowd of people who have closed contact with the subject. Blind spot is the limitation of the surveillance camera system. We also implement deep learning techniques to predict the direction and footage of the subject when the subject enters the blind spot. This technique can avoid the misdetection of the person who might have closed contact with the COVID-19 carrier in the blind spot area.

The Deep Learning Detection of Potentially Infectious Disease (DLDPID) system is used to perform tracking and recognising any potential infectious person in an effective way. With DLDPID, we can help the government to track and check the potential infectious person based on the previous place the infected person went, and the closed contact person to prevent the disease from spreading. This also can raise personal precaution by notifying them about the infected person. 

The Deep Learning Detection of Potentially Infectious Disease (DLIDPI) system can be used to perform tracking and recognising potential infectious persons in an effective way. With DLIDPI, we can help the government to track and check the potential infectious person based on the previous place the infected person went, and the closed contact person to prevent the disease from spreading.

he current coronavirus 2019 disease (COVID-19) pandemic has greatly changed our perspective of the risk for infection from contact. Without any supervision, it can be hard to track the visited place and closed contact person with the potential cluster COVID-19. Therefore, surveillance cameras play an important role to supervise any potential cluster to prevent the spread of disease. The DLDPID system can predict the path that will be taken by the user if the user was not captured by the surveillance camera. With the help of deep learning, this system can identify and track a person, multi-person or people correctly. It can monitor continuously 24 hours per week without having a rest. In this way, the possibility of time taken to detect potential clusters can be reduced and prevent the spread of COVID-19.

Over the next one year, our project can assist in identifying the COVID-19 cluster. The Deep Learning Detection of Potentially Infectious Disease (DLDPID) system  can effectively identify the COVID-19 cluster through the use of surveillance cameras with cloud services and deep learning. DLDPID system can reduce the manpower to monitor the camera continuously 24/7 without resting. Hence, the manpower can be fully utilized to serve other purposes. In the next three years, we can introduce this project to those who need to detect any potential disease infector. It can act as a cluster detection and security purpose. It can notify the user about the person who passed by or made contact with a person. This can reduce the spread of the COVID-19. Moreover, this project can be commercialized or provide services

First of all, artificial intelligence must obtain a high accuracy (above 80%) on tracking and recognition compared. It can record the location and times of a person previously went. Moreover, it can also generate information of the close contact person with the infected person.

Next, the success against the impact goals can also be measured by the reducing rate of infected people within the area of surveillance. The aim of the system is to prevent the occurrence of infectious disease crowds.

Thirdly, it can also be measured by the number of potential clients who demand for the system.  For example, the kindergarten where most of the children will go. By employing this system, it can prevent them from getting infectious diseases such as COVID-19. The DLDPID system can detect each child's temperature. Besides, it can also act as security camera to prevent any stranger go into the place and kidnap the children without anyone notice

Barriers 1: Financial

Since we are doing research on deep learning algorithms and cameras, we need to purchase a few computers that can train deep learning, hardware and components for cameras. Thus, the purchasing cost is huge.

Overcome 1:

We need to have financial aid in order to have enough cash for buying the material we need to carry out research projects. With enough materials, we can obtain more data to enhance  results.

Barriers 2: Facility

Since our project involves usage of GPU and computer, a facility that provides controlled conditions (clean environment, personal protective equipment provided) and good security is needed.

Overcome 2:

We can find and request a laboratory suitable for research, but first we need to have sufficient funds to rent the facility.

We apply to join The Trinity Challenge because our interest is in line to its objective in preventing and reducing pandemics such as COVID-19. Moreover, the number of covid clusters has increased every day which could affect the economy worldwide as well as Malaysia. With Trinity Challenge’s funding, we could build a Deep Learning Detection of Potentially Infectious Disease (DLDPID) system to detect potential infectious disease through surveillance cameras. This can help to protect the kindergarten child from being affected by Covid-19. We have a strong fundamental on our solution whereby we can give an impact on the community and help to reduce COVID-19.