Bench to Bench
An open-source accessible walking route planning application designed for people with limited mobility, such as those with knee pain. It combines OpenStreetMap data with a local visual AI model to plan circular walking routes that guarantee a seating option within every 150 metres and maintain smooth surfaces throughout. The application also supports voice navigation.
Timeline (last 90 days)
Developer Koushik built the open-source project Bench to Bench for his father—a walking navigation app that plans short circular routes around the user's current location, guaranteeing a place to sit no more than every 150 meters and avoiding steps, gravel, and cobblestone surfaces.
In the Mainz Neustadt district in Germany, OpenStreetMap has mapped 431 benches, and the app uses this data to plan multiple circular routes of 590 meters, 987 meters, and 1294 meters, with the longest unseated segment being only 149 meters.
The core architectural principle of Bench to Bench is that Gemma only describes photo content, while all routing decisions are handled by deterministic code.
The author acknowledges that no one has actually walked a route yet, and all testing was completed at home using simulated walking.
All Facts (4)
Developer Koushik built the open-source project Bench to Bench for his father—a walking navigation app that plans short circular routes around the user's current location, guaranteeing a place to sit no more than every 150 meters and avoiding steps, gravel, and cobblestone surfaces.
50%UnverifiedIn the Mainz Neustadt district in Germany, OpenStreetMap has mapped 431 benches, and the app uses this data to plan multiple circular routes of 590 meters, 987 meters, and 1294 meters, with the longest unseated segment being only 149 meters.
50%UnverifiedThe core architectural principle of Bench to Bench is that Gemma only describes photo content, while all routing decisions are handled by deterministic code.
50%UnverifiedThe author acknowledges that no one has actually walked a route yet, and all testing was completed at home using simulated walking.
50%