Humpback Whale Identification Challenge

https://www.kaggle.com/c/whale-categorization-playground

Python Environment Setup

Set up a conda environment for the project

1. open this project in PyCharm
2. open up settings
3. select Project:Whale/Project Interpreter
4. click on the gear
5. select conda, new environment with python 3.6
6. click ok

Data Setup

1. in Whale root, mkdir data
2. download the files from https://www.kaggle.com/c/whale-categorization-playground/data into data
3. unzip train.zip and test.zip
4. move the train dir to raw train: mv train raw_train
5. run the src/utils/prepare_data.py script (you can right click on it and click run)

• this will create a directory in data called train which will have images in the format of label/example1.jpg

Running tests in Pycharm

1. click on Edit Configurations: from the run menu
2. click the + button, and then Python tests/Unittests
3. name it something like unit tests
4. for the Target/script path navigate to Whale/tests
5. under pattern, put *test.py
6. click on ok
7. click on the play button to run the tests

Running tests from the command line

1. create a conda environment using the requirements.txt: conda install --file requirements.txt
2. run the following command (outputs to TestOutput/log and git adds it): ./run-tests.sh

Training and Predicting

Either:

• right click on trainer.py in PyCharm and click the run button
• run from the command line: PYTHONPATH=~/PycharmProjects/Whale python src/whale/trainer.py