AccessAu

Inspiration

Last school year, one of our friends had several hip surgeries and subsequently needed a ride to campus every day. While on these drives, we started talking about how inaccessible Golden is. There are sidewalks that have stairs on them (with no ramps), steep hills, and gravel walkways present around campus and downtown. Google Maps has a (very hidden) feature to avoid stairs, but it is incorrect somewhat frequently and completely disregards steep inclines. We decided to build this app to help Golden become more accessible by allowing people to customize their route finding experience!

What it does

Finds the best path based on user preferences (1-10 for sensitivity to steep inclines, as well as a boolean to avoid obstacles like stairs or gravel paths). Provides step by step directions as well.

How we built it

We used Google's Elevation and Streetview APIs to provide elevation data for each node in the graph (used to weight edges according to user's incline sensitivity) and get street-level images of the paths people would take. The latter was crucial to our application, as it allowed us to send images back to Groq and get coordinates

Challenges we ran into

Learning how to use OSM was a bit challenging, as it took a second to realize we had gotten the filter wrong. Additionally, getting the vision model to just return a binary classification required a lot of tinkering with the prompt. Finally, we couldn't get some of the street names to populate correctly in the directions when travelling on a sidewalk parallel to a street (the graph didn't contain roads that were strictly for driving with no room for pedestrian traffic), so we had to call those "sidewalk"

Accomplishments that we're proud of

We were so incredibly tired when it came time to add the directions. However, we managed to get it to work pretty well given our fatigue, and I'm glad we were able to provide more specific instruction to the user.

What we learned

Always plan ahead. We got to the very basic terminal interaction version of this pretty quickly, and were stuck scratching our heads on what the best order to do subsequent tasks would be.