A leading Baltic frozen and convenience food producer Mantinga introduced Computer Vision powered solution to ensure PPE compliance. During a three-month pilot project, the AI PPE detection product was installed at the company’s logistics center and a manufacturing plant that is currently being built.
The PPE detection solution was tasked with monitoring whether employees are wearing the required protective equipment, including hardhats, high visibility vests, and face masks.
The AI-based PPE solution analyses video footage from surveillance cameras. The solution automatically identifies employees and PPE. It is able to notify the responsible worksite safety officer in real-time about PPE policy breaches. The solution also produces periodical reports on identified PPE policy breaches. The solution is fully compliant with EU GDPR regulations.
During the pilot project, high visibility vests and hardhats were identified with 93% accuracy, while the accuracy for face mask detection was 100%.
The PPE detection solution was also trialed at the building site of the new manufacturing plant. In 2021 Mantinga plans to invest 21M Eur to increase and improve the company’s manufacturing capabilities.
“During the pilot project, a few different contractors were working on site. The employees of every contractor were wearing different types of helmets and vests. This called for additional training of our machine learning model. However, the accuracy we achieved is a good indicator for what could be expected from AI PPE solutions in real-world conditions – when you have changes in lightning, objects may obstruct the view for the cameras, etc. All in all, the AI PPE assistants can clearly supplement the compliance monitoring by company employees” – noted Head of Products at Agmis Simas Jokubauskas.
Simas Jokubauskas, CEO of EasyFlow
During the pilot project at Mantinga, in addition to PPE compliance monitoring, the AI worksite assistant was also tasked to identify work process anomalies. The solution would monitor whether forklift operators adhere to safety protocols, identify any damage to cargo when it is being loaded or unloaded.
“Such monitoring can be undertaken by utilizing the same surveillance camera infrastructure already present on the company’s premises. If an event can be filmed, in most cases an AI model can be created to automatically identify and quantify such events” – added Simas Jokubauskas.