Computer Vision in the Time of the Coronavirus Outbreak

Mark Altaweel

Updated:

New tools and older ones used for very different purposes are emerging after the coronavirus (COVID-19) outbreak. One interesting trend is the use of computer vision tools and techniques to improve monitoring of urban spaces and medical data used to diagnose COVID-19. While such tools were used for safety or traffic monitoring before, they now have a new function in this crisis.

Using Computer Vision to Map How People are Using Urban Space and Practicing Social Distancing

One interesting tool is created by Numina, which is a company specializing in monitoring movement and traffic using video data. Previously, the company was using this data to help cities plan traffic, including cars, pedestrians, and others using busy streets. The tools the company has enable a new form of monitoring, mainly measuring social distancing in cities such as New York.

The company even used data on pedestrian movements to see how effective people were in maintaining social distancing (2-meter) guidelines as people navigated the city. In a city like New York, the company highlighted, sidewalks are not always wide enough to maintain this distance, making it difficult to always maintain the guideline.

Multi-object tracking in video enables objects, and their pathways as they move through a space, to be monitoring and focused on, while depth measurements enable measurements to be made. The video in the Tweet below shows that even with the best intentions, as most people tried to maintain a health distance from others, people struggle in New York to maintain their safe distances due to space limitations.[1]

In China, computer vision was used extensively for monitoring people. While such technologies might be worrisome in the West in regards to privacy and state monitoring of people, there are some advantages. For one, Chinese startups SenseTime, Megvii, and DeepGlint developed thermal camera monitoring, which can detect temperatures and using computer vision and artificial intelligence classification, there technologies could detect the likelihood someone walking around was infected with COVID-19.

Furthermore, monitoring tools created by tech giant Baidu was used to monitor if people were wearing masks or not. This enabled authorities to quickly determine if people were violating laws against not wearing a mask in public. Another company, Yitu, used its medical imaging technology to find signs of infection in patients in an automated process, helping doctors overwhelmed in Hubei province.[2]

Deep Learning to Better Classify Coronavirus Results

Research is also emerging where deep learning methods are used to better classify coronavirus results. So far, results are promising using various CT scan methods, where research has shown over 98% sensitivity and 92% specificity in identifying COVID-19.[3] Using CT scans has become recommended for more accurate diagnosis; however, doctors may not always easily distinguish normal pneumonia for COVID-19. Using CT scans, doctors reported specificity ranging between 60–70%, indicating that AI, deep learning methods on image classification could improve the accuracy of diagnosis even when compared to trained professionals.[4] 

Open source tools such as Keras, which uses Google’sTensorFlow library, a popular tool for computer vision and AI, has also recently demonstrated even with simple X-rays, COVID-19 could be detected by AI, deep learning methods. Such a tool is free and open-source, as Google is also encouraging developers to create new and better techniques in helping to identify COVID-19 in patients.[5]

Computer vision is emerging as a promising set of techniques such as enabling rapid detection of COVID-19 and the better monitoring of social distancing practices among people. While these techniques are promising, the scale of the problem likely means most places will not see the benefit of these techniques. However, what it does suggest is that after the crisis subsides, new research is likely to attempt to improve these methods even more and we can expect them to become more prevalent. 

References

[1]    For more on video tracking technology and multi-object detection computer vision, see:  https://numina.co/why/

Also see: Presentation by Numina to the Transit Techies NYC meetup that explains how Numina’s technology works:

YouTube video

[2]    For more on computer vision used in China for various purposes during the COVID—19 outbreak, see: https://go.forrester.com/blogs/computer-vision-joins-the-fight-against-coronavirus/.

[3]    For more on using AI and deep learning on CT images, see:  Gozes, O. et al. Rapid AI Development Cycle for the Coronavirus ( COVID-19 ) Pandemic : Initial Results for Automated Detection & Patient Monitoring using Deep Learning CT Image Analysis Article Type.  Radiol. Artif. Intell. (2020). DOI: arXiv:2003.05037

[4]    For more on doctor assessment of CT images and diagnosis of COVID-19, see:  Bai, H. X. et al. Performance of radiologists in differentiating COVID-19 from viral pneumonia on chest CT. Press Radiol. (2020). https://doi.org/10.1148/radiol.2020200823

[5]    For more on COVID-19 detection using TensorFlow, see: https://www.pyimagesearch.com/2020/03/16/detecting-covid-19-in-x-ray-images-with-keras-tensorflow-and-deep-learning/.  

Related

Photo of author
About the author
Mark Altaweel
Mark Altaweel is a Reader in Near Eastern Archaeology at the Institute of Archaeology, University College London, having held previous appointments and joint appointments at the University of Chicago, University of Alaska, and Argonne National Laboratory. Mark has an undergraduate degree in Anthropology and Masters and PhD degrees from the University of Chicago’s Department of Near Eastern Languages and Civilizations.