Secure computing with few examples

Contrary to popular belief, today’s video surveillance still depends on expensive, daunting and error-prone manual inspection. Automation is challenging because activities of interest are rare, scenes are over-crowded, computing demands are humongous, and offloading to commercial label and compute services is blocked by privacy laws. The project studies these research challenges by exploring AI-tactics that are less demanding in terms of labeled examples; together with new high-performance computing architectures that are scalable and secure in terms of their video processing capabilities. All research results will be integrated in a real-time video surveillance search engine for Amsterdam Schiphol Airport.

Amsterdam Schiphol Airport

At Schiphol more than 3,500 cameras observe the facility. Its surveillance video streams are monitored 24/7 by a dozen security, customs and military police officers. The commercial, government and airport stakeholders in the project, Schiphol Nederland BV, KLM, Royal Netherlands Marechaussee, Customs, Ministry of Internal Affairs and Kingdom Relations, Dutch Railways, Robert Bosch BV and Griffid BV, have identified the need for a large-scale, high-performance video AI system that outperforms humans at a task for which automated industry solutions still provide poor results. To address the challenge of real-time video AI, the project provides funding for three PhD students and a software engineer.