Inspiration
I was trying to rent out an airbnb to watch a couple important sport matches with my friends (ufc & world cup), but holy these prices were crazy. I then checked hotels nearby out of curiosity, and realized that I was being extremely overpriced. I wanted to build some tool to help hosts reasonably price their airbnbs, as well as to help finders like me figure it out if im getting scammed or not.
What it does
Fairbnb allows users to validate listings for an airbnb by comparing it to identical looking hotels. It takes an image or video of an airbnb room and embeds it as a vector. Then it queries stay22 using its direct travel API with location similar to the user's input, and embeds the image of each hotel. Then, using cosine similarity, Fairbnb compares each vector to your image/video vector, and selects the top three most similar. Finally, it averages the price of these 3 identical looking hotels, and presents you a recommended price.
How we built it
Challenges we ran into
Finding a way to embed video data, there isn't much out there. Figuring out how to compare images was hard and took a lot of researching. I had no idea that you can even turn stuff into vectors, so figuring out what embedding did and why it was better than other stuff was challenging.
Accomplishments that we're proud of
A full upload -> embed -> compare -> output pipeline that is seamless to the user, and a polished UI.
What we learned
How vector embeddings work and how they are used for similarity searches.
Log in or sign up for Devpost to join the conversation.