Project Overview
For a while, I was interested in learning more about machine learning, so I created a rudimentary convolutional neural network (CNN) to help me organize my art files based on artwork style. I remodeled the basic Cats vs. Dogs classification ML project to suit my needs. Instead of identifying images between dogs and cats, my model predicts if an image is a line art or a colored image. Then, it sorts the images into their respective folders based on its prediction. One question I always get asked is why I don't just check the color values of the pictures and then sort the images. The issue with this approach is that if the line art were drawn on colored paper, then this method would fail.
The training dataset consisted of around two thousand images scraped from the web, which were split into two groups - one thousand color images and one thousand line art images. The remaining images were used for validation batches.
The attached demo video showcases the general workflow of the machine learning model, as well as some interesting failure cases where it identified some grayscale images as colored images.

Technology Used

Project Demo