Implementing Image.to Image transformation algorithms.
Also, deploying a deep learning model in production.
What it does
There are a couple of algorithms provided
Multi Style Transfer for Images: Allows taking an image from a file, webcam or sharing a screen and applying 16 different styles with different proportions.
Comixifying Images: Take an image similar to the Multi Style Transfer, and comixify it in a manner similar to japanese anime style.
How I built it
The models were first trained on Google Cloud ML Engine, and it is also used as a prediction service.
The web service is implemented in Go, and the frontend uses material components.
Additionally, for Multi Style Transfer, an offline model is also provided for on device prediction, as well as offline prediction.
What's next for stylizer
- more image to image transformation models, see the readme in the tasks folder in the repo