Header Image

RAILIFY @ BaselHack 2024

Be part of shaping the future of rail transportation with Siemens Mobility! Harness the power of AI and machine learning to digitize railway plans, solve real-world problems, and contribute to the advancement of sustainable mobility. Show your skills, innovate, and help revolutionize the industry!

Join the Challenge!

Task of Challenge

As a railway engineer at Siemens Mobility, I want to utilize a digital node-edge model of my railway network. Unfortunately, information gets delivered in old school vectorized PDF and TIFF formats. I’m far too lazy to manually engineer that huge variety of visual data.

Help me to recognize tracks, switches , signals, annotations and other entities and to arrange them in a standardized model for my work. Bring us on the engineering fast track when we e.g. have to modernize interlockings or determine optimized and safe routes through the rail network. This is much more than just image recognition! The real challenge lies in the variability of how plans are drawn and how information is arranged.

My vision is to use the extracted topology data as input for further sophisticated railway solutions. To have a digital model that spans the entire engineering process and life cycle.

Business Engineering

Your tool for extracting existing plans into a digital model is a crucial step in speeding up the engineering of safety relevant interlockings, ETCS radio block centres, or timetable engineering. The collaboratively gained information helps us as to deliver modernized railway solutions to you as potential passenger.

Technology

As customary for a hackathon, the choice of technology for the implementation is completely up to you. Potentially you’ll be working with OCR, image recognition and AI based modelling of typical railway elements.

What you get

Task Image

Expectations

Minimally we expect that you can solve the challenge for the easiest category of railway plans consisting of 2-3 tracks. You’ll earn more respect if you’re able to transform more complex maps containing more variations of elements in the layout.

The more accurate the delivered digital model becomes (rail symbols, correct associated annotations), the easier our rail engineer’s life becomes.

Visualize the elements your tool has been able to digitize into the model.

Evaluation Criteria

Below is a detailed breakdown of the areas there will be focus on during the evaluation process:

  1. Complexity level achieved (L1 to L5)
  2. Completeness of detected symbols (quantity of numbered tracks, switches, signals, axle counters)
  3. Correct modelling of annotations / labels (correct assignment to the edge & direction, elements carry correct properties, relative distances)
  4. Visualisation of model
    1. Layout information can be taken from plan-coordinates
    2. Suggestion: Graphical plan overlay (for verification, highlight what you’ve recognized, indicate also what didn’t work)
    3. Automatic count of elements
  5. On Sunday 11h: you get another L3 plan for processing. Surprise us with your results.
  6. Revisions of one L3 plan: as an extra challenge, we would like to see the difference of plan revisions. Therefore, one L3 plan is available as version B and version C. The differences should be highlighted in the digital model.
  7. Ignore Footer / Header / Metadata from plans

Material to download

Below you cand find all required inputs and useful material for our challenge:

AWS processing power: we can provide your team with AWS vouchers. Please come and see us.

Download all

Winners

Congratulations to Team DerAIler for winning the challenge! Your hard work, innovative approach, and dedication have truly set you apart. Well done on this well-deserved victory!

Railify challenge's winner team: DerAIler

Career

Join Siemens to contribute to the future of technology and AI. Learn more about career opportunities.

Explore Careers