Case Study: TNO achieves week-to-hour traffic-video analysis to boost road safety with Google Cloud Platform

TNO Making roads safer with TensorFlow and deep learning

TNO, the Netherlands Organization for Applied Scientific Research, applies lab‑grade research to real‑world problems and partnered in the InDeV project to improve traffic safety by identifying dangerous urban intersections. Reviewing weeks of CCTV to find the few critical incidents is time‑consuming and hardware‑intensive—typically six weeks of footage requires weeks of manual analysis and many clients lack the specialist GPUs needed for deep learning.

TNO built a TensorFlow/Keras video‑analysis pipeline that localizes pedestrians, cyclists, and vehicles and runs on GPU instances in Google Cloud Platform, letting clients “plug and play” without buying specialist hardware. The cloud‑based solution and open TensorFlow ecosystem sped development and deployment, cutting a week’s manual review down to about one hour, enabling scalable, faster delivery of safety assessments to municipalities.

TNO

Maarten Kruithof

Data Scientist