-
-
The user enters two URLs, one leading to a content image and the other to a style image.
-
After pressing go, the application produces the style transferred image.
What it does
Ever wonder what your dog would look like if Picasso painted her? Me too. This web app, built with a React JS frontend and a Flask backend, prompts the user for two images: a content image and a style image. When the user presses go, the content image is reproduced in the style of the style image. Cool, right?
Inspiration
The first time I saw style transfer in action, I thought it was magical. I wanted to reproduce them myself to gain a better understanding of the technology.
How I built it
React JS frontend, Flask backend
Challenges we ran into
First time implementing a Flask backend on my own, wasn't sure where to begin!
What we learned
Learning the underlying ML concepts in style transfer.
What's next for Air-style
Site styling (frontend), allow user to upload/drag in images, enable user to order a print of their image
Resources
https://towardsdatascience.com/light-on-math-machine-learning-intuitive-guide-to-neural-style-transfer-ef88e46697ee https://www.fritz.ai/style-transfer/ https://genekogan.com/works/style-transfer/ https://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Gatys_Image_Style_Transfer_CVPR_2016_paper.pdf
