Friday 21 April the SimLab centre at TPM organised a hackathon to help ProRail improve safety and efficiency of the Dutch Railways. 

Six teams of 5 persons went ahead to find, scan and report train wagon numbers. These labels indicate whether the wagons cargo contain hazardous materials and therefore provide crucial information for emergency services in case accidents happen. Up until now about 60% of the labels and numbers is being read correctly by the best industrial system. We need the system to be 99% correct, says Scott Cunningham, one of the organisers of the event.

The challenge

Besides students also engineering and computer scientists participated in the hackathon. They were asked to come up with improvements to better identify the labels on moving trains, which proves to be a real challenge for automated systems. The groups worked with real-life data originating from the Smartsensing@Kijfhoek project near Rotterdam harbour, initiated by ProRail, DB Cargo, TU Delft and Railcenter. Each team worked on a different challenge, such as foreground-background discrimination or identifying numbers hidden under rusty wagons or graffiti.

Deep learning

The hackathon provided valuable insight in deep learning procedures, algorithms that help machines act more like the human brain in image recognition. The project partners will use this information to add more sensors to the container of Smartsensing@Kijfhoek and put more smart-sensing test containers in other areas to monitor cargo trains. The winners of the hackathon will be rewarded with an all-expenses paid railway weekend to a railyard in Germany.

Image credits – Jan Sluijter photography and Big Data Republic