Kaggle Cdiscount's Image Classification Challenge solution with Keras

My code for Cdiscount's Image Classification Challenge. Tested on a subset of 10k samples from the ~7m on the full dataset.

Requirements

In params.py set base_dir to your working directory. You can also set the model to use and training parameters.
Place train.bson and test.bson in {work_dir}/input.
Run read_data_train.py and read_data_test.py to read and unpack train data and test data respectively.
Run train.py to train the model and test_submit.py to predict on test data and generate submission file.