For a while, I was interested in learning more about machine learning. As a result, I created a rudimentary convolutional neural network (CNN) to help me organize my art files base on artwork style.
The "Hello World" program equivalent of machine learning is the dogs versus cats identification project. I borrowed the idea and modified it for personal use. Instead of identifying images between dogs and cats, the model will predict if an image is a line art or a colored image. Finally, the program will sort the images into their respective folders base on its prediction.
The training dataset was around two thousand images scraped from the web. Then the images were split into a group of three. There were one thousand color images and one thousand line art images and the rest became validation batches.
The below demo video showcases the general workflow of the machine learning model and some interesting failure cases.