This session explores a practical AI solution developed by NS and KickstartAI to enhance social safety at train stations by detecting unattended objects such as bags and suitcases. Stations are equipped with many cameras, but continuously monitoring all of them manually is not realistic. This made the use case a natural fit for computer vision.

The talk will dive into how this specific use case was identified within the broader context of public safety, and how the project progressed from idea to implementation. It will address the technical complexities of using state-of-the-art vision models in a low-latency environment, as well as the organizational and ethical challenges of working with privacy-sensitive video data. Attendees will gain insight into what it takes to responsibly deploy AI in a high-stakes, operational public setting and how this type of collaboration can lay the groundwork for broader transformation within the public domain.