This project explores the DeepLearning4J implementation within Scala to classify images for the Yelp Photo Classification problem on Kaggle. It is a multi-label classification problem where each entity can belong to multiple classes. The goal of this project was to experiment with a data science problem in Scala utilizing a deep learning library. It could be possible to train meaningful CNNs with this approach. However, the CNNs produced in the results folder of this repo are by no means novel, as I currently don't have the patience or resources (GPU, EC2) to train these models at scale. My intentions are pedantic, so this is more of an experiment/tutorial than a shot at the non-monetary Kaggle grand prize.
/src/main/scala/modeling/main.scala
is the code to run the project end-to-end. Training CNNs is very time consuming, so it is
likely that you will not often run the project in its entireity very often.
Sbt
run