Applying machine learning to keep employees safe and save lives

Presented by AWS Machine Learning

Whether on factory floors, construction sites, or warehouses, accidents have been an ongoing, and sometimes deadly, factor across industries. Add in the pandemic — and an increasing rate and intensity of natural disasters — and the safety of employees and citizens becomes more complicated.

Australian-based Bigmate, a computer vision company focused on enhancing workplace safety, is using machine learning to reduce workplace accidents, help companies detect potentially ill employees as they arrive on site, and aid organizations in the operational management of natural disasters.

Bigmate’s risk management and computer vision expertise combined with their long-term experience in asset management are all supported by their in-depth knowledge of advanced AWS Services to maximize operational turnaround.

“Organizations are deeply concerned about safety, and are looking to what AI and ML can bring to the table, not for the sake of technology but to help improve safety in the workplace through targeted applications with clear benefits.” says Brett Orr, General Manager at Bigmate. “Our engineers’ superpower is using computer vision to identify unsafe situations, and pairing that information with existing and new sensors, to understand where things are, what they’re doing, and if they’re working as they should.”

Reducing accidents with a high degree of accuracy

Work-related injuries cost the economy $61.8 billion with the costs borne by both the organization and the worker themselves, making this extremely challenging for both parties. Many of those injuries happen on the factory floor which is over-represented as a proportion of all work-related injuries, Orr says.

“70% of workplace injuries or deaths happen because of unwanted interactions between heavy vehicles and people,” he says. “Of those 70%, more than 30% are unwanted interactions between forklifts and people.”

Organizations have stepped up their Occupational Health & Safety activities to make employees more aware of workplace dangers and improved safety measures, which does help, but accidents can still occur.

Bigmate developed Warny to enhance safety in the workplace and reduce these kinds of accidents. The Warny ecosystem is comprised of three core applications: vehicle collision avoidance, safety zone alerting, and thermal analysis of people and industrial systems on the factory floor.

Developed over a number of years, the solution is built in-house and uses both edge hardware for local performance and privacy as well as AWS Services in the cloud. Warny leverages many of the key AWS Services such as IoT, Greengrass, and Sagemaker.

Warny uses sophisticated computer vision algorithms to protect people working around dangerous machines — such as forklifts, trucks, or manufacturing machinery. It can detect instances of spontaneous combustion of materials, overheating of equipment, and fires in the workplace, as well as analyze, report on, and alert machine operators in real time about unexpected events, such as a person being in an unsafe area even when not in line of sight of the operator.

Early into installation, one factory in Singapore saw a 22% drop in incidents, an impressive precursor of what was to come when completely installed. Across the board, on average, Bigmate is seeing a 80% drop in incidents in their clients’ work environment.

Using edge-based software and cloud-based services, data is streamed from IoT sensors to the platform, which analyses images and object detection data. They’ve created a neural net that can recognize equipment and people with near-100% accuracy, says Orr.

A significant portion of the solution is the ability to create depth of field on a standard CCTV camera. With that depth information, they can determine position, distance between objects, and speed to begin to audit the environment, understanding the movements of people and forklifts.

Using trajectory data calculated from depth, position, time, and distance, machine learning algorithms then predict an object’s path in near real time. If a collision is imminent, the platform sends out an alert to the employee’s wearable, with only milliseconds of latency thanks to AWS IoT Greengrass. The data is then collected in dashboards to allow organizations to analyse their workplace safety practices.

Using AWS Services, Bigmate was able to develop their platform in a little over nine months with a small core development team, and a 15-20% reduction in development resources overall.

“For us, the partnership with AWS is not just important. It’s critical to what we do,” says Orr. “AWS has allowed us to scale infinitely and quickly, allowing us the flexibility to turn applications on a dime, and the ability to work with industry best practices inherent in the products that AWS sends out.”

Making it safe to go back to work

In a pandemic, the choice between returning to work and staying safe has been particularly challenging. Now, safety measures such as mask and social distancing policies, as well as temperature screening, are allowing businesses to reopen their doors and keep their employees healthy.

Manual temperature screening, however, can be a blow to productivity, causing bottlenecks at entrances and exits, and it doesn’t catch employees who begin to get sick during the day.

Bigmate’s pre-screening solution, Thermy, tackles those issues, using thermal imaging that can immediately detect elevated temperatures of people in real time, at scale, scanning 30 people a second, 500-600 people a minute, plus run 8.3 scans per second to validate its readings.

It can be deployed in any location in an organization — at the entrance of the building or the factory floor, cafeterias, breakrooms, washrooms, and anywhere else employees move throughout their day.

The solution, which is based on the Warny platform and technologies, uses thermal cameras and advanced analytics with machine learning, providing real-time information through dashboards hosted on AWS for remote viewing and trend analysis.

Other thermal solutions only capture skin temperature, which doesn’t accurately diagnose core body temperature. Bigmate’s platform calculates a true representation of a person’s core temperature. It first uses computer vision technology and the data from a thermal camera and an optical camera to isolate the subject’s head, to capture skin temperature even when the subject has a beard, glasses, a hard hat, or other features. A machine learning algorithm can then calculate a representation of the core body temperature, to determine whether the subject has a temperature.

The product was originally designed to help stop the spread of flu and other highly contagious diseases being spread throughout an organization. When the pandemic struck, Bigmate was able to deploy the solution immediately to organizations concerned with keeping their workers safe from COVID-19.

“In organizations that have had 50% or 60% positive rates, it’s helping to reduce the spread and hot spots before they happen,” Orr says. “It means businesses can continue to run.”

Mitigating natural disasters with ML

Bigmate also leverages its computer vision and thermal detection technologies to help state and federal government organizations detect natural disasters, from floods to tsunamis to forest fires, in real time.

They use imaging taken by high-tech cameras sitting on fixed-wing aircraft and helicopters to capture data like the latitude, longitude, time, and the distance a disaster is from landmarks. The technology also calculates information about the chopper’s height, speed, and the focus point of the cameras to understand exactly where the incident is, geographically.

They send the real-time metadata to the bureau of meteorology in order to factor in weather data, such as windshear, wind direction, rain, and so on. That data is merged with population and location data.

This gives safety and environmental organizations that respond to disasters the information they need to project where resources should be deployed to save lives and property safely and effectively — whether that be to quell a fire, clear out a campground, prepare for worsening conditions, and more.

The social promise of ML and computer vision

The promise of machine learning goes beyond business challenges — solutions like these are a demonstration of the impact technology can have on society, tackling health, environment, and safety issues that were previously difficult without the help of AI, machine learning, and new innovation.

And at Bigmate, the work they do makes a difference every day, Orr says. They’ve delivered some important milestones, from the number of accidents their machine learning algorithms have been able to prevent, to the number of outbreaks their computer vision technology has helped to reduce, to the lives they’ve helped emergency services save.

“A lot of the time you can’t point back and see where the work that you’ve done has had a direct impact on people and their lives and families, but we’re able to bring our experience and technology to bear to getting people home safely,” Orr says. “That’s a big one.”

Dig deeper: See more ways machine learning is being used to tackle today’s biggest social, humanitarian, and environmental challenges. 

Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. Content produced by our editorial team is never influenced by advertisers or sponsors in any way. For more information, contact [email protected].

Source: Read Full Article