I built a convolutional neural network on the basis of ResNet50 with the goal of classifying images of used lego sets from three different product lines: Duplo, Star Wars and Technic. I am using the data generated with my Ricardo webscraping code (see my other repository).
The network currently achieves a classification accuracy of around 75%. It is work in progress. You can check out all the code in the Jupyter Notebook.
Note: This repository was inspired by an in-class exercise during my Data Science Bootcamp at Propulsion Academy.