Google's DeepDream neural network is around for quite some time now. Although it creates fantastic psychedelic images, it still lacks the best part of human dreaming: the rapid eye movements during the so-called REM sleep. So why not combine human eye interaction with DeepDream somehow?
What it does
This project aims to combine DeepDream's capabilities with the actual eye movement of the user and tries to generate some cool art. To accomplish this, the web application tracks your eye focus point from a webcam and maps it onto the image. We are then able to create clusters of the most interesting regions in the image. These regions are put together into a collage and then fed to DeepDream.
How we built it
The application basically consists of a website with a python back-end. Libraries used include: webgazer.js, flask and deepdreamer.py
Challenges we ran into
None of us is a webdeveloper, so building a website was quite time consuming. Apart from that: Bugs, bugs, bugs....
Accomplishments that we're proud of
We were able to track the eye movement in a nice-looking, client-side webapplication!
What we learned
Basic Webdesign REST-API Backend written with Flask
What's next for DreamGazer
There are still a lot of bugs and the DeepDream integration is not finished.