Pastiche

Inspiration

This website which maps lines that people draw to random satellite images

What it does

This project maps out contours of a database of artwork and then attempts to match the known artwork patterns to your given gestures.

How we built it

We used a python backend which included Deeplake to access the dataset and pyTorch to edit and filter out the dataset. The frontend was built using Flask, axios and React.

Challenges we ran into

Learning how to integrate the front end with the back end was challenging. Also, trying to figure out how to match the drawing the user gave to any of the contours had us stuck for a while since we had to match the pattern of the drawing, and not just the absolute (x,y) coordinates of the two contours.

Accomplishments that we're proud of

Since everything was new to us, we are proud of how far we ended up coming with no foundation to any of the skills that we now know after having attempted this project.

What we learned

We learned how to use React, Flask, Deeplake, pyTorch, opencv and axios and how to build a functional full stack project. We were all new to some parts of a full stack project so I believe we have all grown further with this endeavour.

What's next for Pastiche

We could include ML models to filter out the time taken to search for matching contours drastically. We can also implement client-side matching for faster results or use Google Cloud to host the dataset so that we have a bigger collection to work with and aren't limited by our local storage options.