Inspiration
We fell in love with playing GeoGuessr from the time we found out about it and it quickly became something that we enjoyed frequently. We really wanted to create something that would help bolster our skills by helping us to learn what to look for with different locations
What it does
We trained an AI to recognize characteristics of different locations in order to recognize those locations from the images we capture from active GeoGuessr games
How we built it
We trained an AI to recognize different locations by scraping images from google maps. We then used selenium to capture the images from active GeoGuessr sessions and passed those through our model to obtain geolocation information. We then return this information to the user so that they can find a pinpoint location
Challenges we ran into
We ran into a lot of issues with our initial data scraping and processing. When we finally got that working, we were able to effectively train our AI. On top of that, we also had issues with using the google maps API to return a map location based on the coordinates, and unfortunately we had to settle on just returning the coordinates to the user
Accomplishments that we're proud of
We are proud to say that our AI was trained based solely on images and information that we scraped ourselves. We are also proud to say that we created our own Chrome Extension that effectively communicates information between a local server, our model, and the extension
What we learned
We learned a lot. We learned how to use selenium webdriver, front-end programming with js and html, and being able to create a server to run a script to get everything to work together
What's next for GeoGuessd
We would like for the program to use the coordinates received from the model to display a specific location on a google map display
Log in or sign up for Devpost to join the conversation.