This project trains a neural network to recognize the structured wayfinding system found throughout train stations, making this hidden layer of information accessible to blind and low-visioned travelers via a smartphone app. Pictograms, numbers, and words on signs are intelligently filtered and semantically enriched to provide only the information relevant to the user’s current goal (e.g., platform 12). This untapped resource complements existing tactile systems like guidance strips and Braille signage, offering improved orientation – especially in unfamiliar stations. A key challenge is mapping of recognized signs to their spatial context andmaking the information truly usable for the target group.