The goal of the project is to improve a kaggle project about Dog Breed Classification, achieving an higher test accuracy. The original project achieved 79% of accuracy on the test set, while this one goes up to 87%. Also further improvements were made to the data processing pipeline in terms of modularity and performance.