In this project I create a CNN to clasify lego images from a kaggle dataset. These lego peices were rendered in 3D using Blender. I use data augmentation with a custom built CNN and Keras Callbacks to acheive a 96% accuracy. I used Kaggle's cloud GPUs to run my model and load my data.
https://www.kaggle.com/joosthazelzet/lego-brick-images
Contains 6400 images of 16 different Lego bricks. The bricks are classified by folders. The images are computer rendered using Blender. For more info please contact me.
