We all have an interest in the applications of Machine Learning, and were on the Tech + Research track for Technica!
What it does
This project allows you to transfer the style of one image onto another.
The style transfer gets some images, applies the style of some art or pattern from other images, and makes an output image; transform the input image using the style from other images.
In the report we based on by Johnson, Justin, "Perceptual Losses for Real-Time Style Transfer and Super-Resolution", they use the style as an artistic style in a certain picture, and content for the image wants to apply those extracted styles.
For example, if we set the style image is 'The Starry Night' from Vincent van Gogh and set the target image as a base image, the expected result is the base image expressed by The Starry Night's coloring pattern, textual expression, and other artistic features.
How we built it
We built this project using Python with PyTorch and on Google Colab notebooks.
Challenges we ran into
We ran into some issues when implementing our training network, but it worked out in the end!
Accomplishments that we're proud of
We're proud of the results of our style transfer network which includes transformed images in several different styles!
What we learned
This project taught us more about the implementation of neural networks as well as image processing and style transfer.
What's next for How AI is making your smartphone a better camera
Next, we plan on looking into more recent efforts in image style transfer and exploring new methodologies. Maybe even look into transferring multiple images or videos.