#Japan #AI Agent #Workplace Safety #Fujitsu #Kawasaki

Fujitsu Develops an Innovative Video Analytics AI Agent

Fujitsu has made headlines with the recent unveiling of its cutting-edge video analytics AI agent, designed specifically for frontline workplaces. This sophisticated AI application aims to enhance both safety and efficiency by utilizing spatial video data combined with workplace imagery and text attributes.

Advanced Technology for Enhanced Safety

The AI is structured on a multimodal large language model (LLM), which allows it to train autonomously. By analyzing 3D images captured from workplace cameras alongside written safety documentation, the AI agent can draft detailed reports and provide actionable recommendations for improving workplace environments. This integration ensures that the AI has the context needed to interpret video feeds effectively while distinguishing significant safety-related behaviors.

World-Leading Processing Capabilities

A standout feature of Fujitsu’s AI agent is its extraordinary capability to process long-duration videos without sacrificing accuracy. By implementing innovative context memory technology, the AI can store relevant details from both video footage and pertinent written information. Users can set prompts within the system, focusing on specific behaviors, such as identifying 'safe behavior in humans'. This bespoke approach to monitoring ensures that safety protocols are consistently upheld.

Collaboration with Academia

Fujitsu is not alone in this venture. It has partnered with Carnegie Mellon University to oversee the development and evaluation of its AI technology. The collaborative project includes an evaluation environment known as FieldWorkArena, which will serve not only as a testbed but also as a resource for researchers in the AI field. This innovative platform will host a collection of videos and images from varied frontline environments like manufacturing plants and warehouses, orchestrated to refine and enhance the performance of the AI agent.

Future Implementation Plans

Fujitsu is set to roll out a trial for this AI agent in fiscal year 2024, shifting towards full in-house application by January 2025. The organization is primed for a major evolution in how workplace monitoring is conducted, with the potential to drastically reduce incidents while boosting operational efficiency.

Moreover, a library of tasks and scenarios will be posted on GitHub as part of Open Source contributions to the research community, providing a transparent means for researchers to interact with and evaluate Fujitsu's advancements.

Conclusion

Fujitsu’s commitment to harnessing the power of AI for workplace safety is unwavering. With its latest video analytics AI agent, the company is poised to redefine the standards for frontline workplaces, ensuring that they remain not only productive but also safe for employees. As Fujitsu embarks on this journey, it is set to make significant strides in how industries approach the domains of safety and efficiency.