I got back from a fantastic holiday with some friends and I wanted one combined photo album which had all of the best pictures. But between us we took hundreds of photos! And I didn't want to have sort through them manually.

What it does

MRI Scanner identifies similar images in a photo album and let's you quickly pick which one you like for your final album. If you don't have time for that, MRI Scanner will do all of that for you and create an album of unique photos. Once you have your final album, you can also search images based on what's in them. Only want an album with people in them? No problem!

How we built it

A Python backend handles image processing, the image similarity algorithm and calls to the Clarifai API for image tagging. Our algorithm computes the Spatial Envelope or "scene" for each photograph and uses this to identify similar images. The backend was integrated with the front end using the Django framework. HTML 5 and Javascript used to power the front end.

Challenges we ran into

Computing image similarity is computationally expensive and takes too long at the moment. Making the page interactive (image selection, tag selection, image carousel).

Accomplishments that we're proud of

Getting our image similarity algorithm to work well. Integrating image tagging and tag search functionality. Building an image carousel.

What we learned

How to compute image similarity, how to make calls to an API, how to use the Clarifai Api, manipulating filenames using the os module, image processing. Using more complex web dev tools. Improving at troubleshooting.

What's next for MRI Scanner

Fully integrating the back and front end, moving it from local host to a server, making it possible for users to upload their own photo albums

Share this project: