After the first day at Junction we wanted to watch a movie together. And it was a real problem finding interesting one. So we decided to solve this problem instead of watching it.
What it does
Using hilarious high-level search filters: 1) number of gunshots, 2) ratio of smiles in a frame, 3) day or night domination, and 4) number of love scenes, you can choose a movie for evening in a fun way.
How we built it
There were 2 big parts. 1) Video analysis. We managed to process top-100 IMDB movies (not just metadata, but videostream!) using modern deep learning technics like face detection, emotion classifier and other. 2) Android app. Our team created amazing application from the UX perspective. It is beautiful, handy and very interesting to use. It's like a game, but with end-use benefits.
Challenges we ran into
It was a challenging problem to handle more than 300Gb of videodata during hackathon. Also it was really difficult to make up interfaces for such new task.
Accomplishments that we're proud of
We're proud of our small and neat app you want to play more and more. Help us! We can't stop!
What we learned
Big data technics in the wild. Advanced sprite animation. Our teamwork now is on the next level.
What's next for Chuzi
Adding 1k movies to our database. Creating new great search filters like: choosing movie by color, filtering by character's clothes and so on.