PROJECT GOAL

Increasing workers safety at industrial mega-sites with use of computer vision

THE CHALLENGE

In the past year, the number of accidents and deaths from industrial incidents has grown by 30%. A big cause of accidents is the incorrect usage of Personal Protective Equipment (PPE).

Our challenge was to research and develop solutions that stimulate the correct usage of Personal Protective Equipment and reduce the number of incidents on industrial sites. Creating a safer work environment for all staff.

THE OUTCOME

We developed a computer vision tool, Theia, that recognises incorrect use of personal protective equipment at industrial mega-sites.

Developed from human-machine collaboration principles, Theia doesn't replace safety supervisors, but supports them. Together, Theia and safety supervisors enable increased workers' safety at dangerous industrial mega-sites.

CLIENT
Total Safety
INDUSTRY
Industrial safety services
TECHNOLOGIES
· object detection · machine learning · edge tpu computing
METHODS
· qualitative research · expert interviews · product design · agile development · pilot project · concept validation
METHOD KIT

// our process

We started with on-site shadowing research and interviews to deepen our understanding of the current safety supervision process at mega-sites. We saw that supervisors are often looking at dozens of screens at a time, which makes it challenging to spot incorrect individual usage of Personal Protective Equipment. That’s how we identified the opportunity to support safety supervisors in monitoring industrial sites.

Based on this research, we shaped the concept for a safety supervision solution that recognises the (in)correct use of Personal Protective Equipment, and alerts the safety supervisor. The system supports safety supervisors and makes them more effective by improving compliance with PPE regulations and reducing accidents as a result.

We collaborated with the Head of Innovation and on-site Total Safety staff to identify the requirements, understand the context and train a computer vision model to recognise (in)correct usage of Personal Protective Equipment. As the internet connection at industrial sites can be spotty at times, we developed the solution using edge computing. The solution was then piloted, tested and refined to evaluate the effectiveness on-site.

The process from concept to pilot project took 12 weeks in total.

// the results

In 2020, Project Theia has won a Silver Spin Award in the category "AI and Voice".
During the on-site pilot phase, the model accuracy in recognising PPE equipment reached 90% and showed PPE violation reductions of 13,4%.
The end-to-end process from research to on-site pilot project was completed in 12 weeks.
"
Working with DEUS on Project Theia has demonstrated that AI has immense power in supporting our safety experts in identifying dangerous situations. At the end of the day, it's all about getting people home safely to their families, and we are leveraging this technology to make this happen.

Dusan Rakic

Director of Innovation and Technology
@ Total Safety Europe/APAC

"
FESTINA LENTE

Curious to learn more?

Reach out to us!