AI to Track Social Distancing – A technology snapshot

The COVID-19 pandemic has brought about disruption across all sectors worldwide. Though a third of the world went through partial or complete lockdowns to curb the spread of the virus, some industries could not transition to a completely remote working model. Sectors like healthcare, food & beverage manufacturing, and call centers needed to have their employees physically present in their workplaces to avoid any sort of disruption.

Call centers were severely affected due to the complicated infrastructure already present in the office. As the situation gradually stabilized, organizations were looking at resuming work in phases. However, the main challenge during this time was to ensure that all employees were following the COVID-specific instructions from the government and health officials. Some of them were temperature checking at the entrance of the office premises, wearing face masks, washing, and sanitizing hands, and strictly adhering to social distancing norms of staying 3 feet apart.

Though norms like temperature checking and mask detection were daily easy to implement, it was challenging for HR executives to monitor the distance between employees constantly. Some offices used CCTV cameras to manually parse daily footage and trace employees who defied distancing norms. However, this method required a dedicated person to go through each video file and record details of the rule-breakers, which was a tedious and time-consuming process, especially in areas with high footfall like hallways, pantries, entry, and exit points.

To overcome this bottleneck, an intelligent AI tool was proposed that could monitor employees at a workplace and identify if social distancing standards are being followed at all times. This cutting-edge solution could receive input from a real-time CCTV feed or a pre-recorded video file. Using advanced machine learning methodologies, the tool could be used to scan the video feed in real-time and note any possible breaches in a centralized database. When a breach was identified, it could take a screengrab of the video and email it to the HR department with accurate details like the date and time of the breach, unique camera ID number. This way the person not following the norms could be identified accurately and be advised against such practices in the future. The solution also had an alert mechanism that was triggered when two people come too close to each other. After consistently using this AI solution for two weeks, the HR department noticed that the number of employees flouting the social distancing norms had reduced drastically.

The underlying technologies used in this AI solution were:

Human Detection: With the video frame as the input, it could accurately identify humans with relatively high accuracy. We used a grid mapping to make it easier to identify which part of the image had humans. The model was constantly trained with datasets so that the accuracy improved with time. Once the human(s) were identified, we added a bounding box predictor feature so that the output would have more clarity.

Distance calculation: The distance between two individuals was calculated using an algorithm, which builds multiple decision trees and merges them together to get a more accurate and stable prediction. It operates by constructing a multitude of decision trees at training time and outputting the class that is the mode of the classes/classification or mean prediction/regression of the individual trees.

This tool can play a crucial role in ensuring that the workplace is safe for employees that need to work on the premises. The nature of the COVID-19 is unpredictable and it can often be a cause for panic. As the world navigates through these unchartered waters, company leaders need to formulate a strategic plan of action for their employees to ensure their well-being. AI applications like these can ensure that employees’ safety is constantly monitored in the workplace.