Watch the Replay: AI, Computer Vision, and the Dashboard at the Port of Marseille
Maritime ports are the beating heart of global supply chains, but managing the unpredictable flow of thousands of freight vehicles every day is a delicate balancing act. When the Grand Port Maritime de Marseille (GPMM)launched the Smart Port Challenge, they posed a pivotal question to the tech ecosystem: How can we use our existing camera network to predict and eliminate traffic bottlenecks?
In our latest webinar replay, Euranova lead project manager Juliette Spina takes the virtual stage alongside co-innovators to show how the answer was built.
Developed in partnership with the GPMM, the project demonstrates how Deep Learning can transform standard, passive CCTV feeds into a high-level predictive logistics hub.
Moving from "Observation" to a 6-Hour Forecast
Rather than asking human operators to manually monitor the port's ~700 installed cameras, the team deployed Computer Vision (leveraging YOLO architectures) across a 6-camera pilot zone in the port's eastern basins.
The resulting platform doesn't just record what is happening; it acts as a "Traffic Weather Forecast", delivering three layers of intelligence:
Extraction: It continuously pulls pure quantitative data—counting vehicles, classifying transport types, and mapping precise velocities.
Detection: By establishing customized density-to-speed thresholds, the algorithm instantly flags anomalous congestion the second a buildup begins.
Prediction: Feeding those metrics into Random Forest machine learning models, the dashboard generates a reliable traffic forecast up to 6 hours into the future, allowing port controllers to proactively reroute logistics before a gridlock occurs.
Solving the Privacy Catch
Deploying cameras in a massive public-transit hub usually triggers the legal reality of GDPR. Because the cameras capture public transit, license plates and human faces are unavoidable. The pipeline was designed to perform rigorous video blurring and image encryption at the source before any spatial analytics were run, respecting GDPR and privacy regulations.
Watch the Deep-Dive Whether you are a Supply Chain Director looking at throughput optimization, a Machine Learning engineer curious about lightweight YOLO deployment, or a Data Protection Officer studying real-time video anonymization, this session bridges the gap between theoretical AI and real-world logistics.
🔗 Click here to watch Juliette present the project on YouTube (jump straight to 10:17)