/databowl2018

Primary LanguagePythonMIT LicenseMIT

databowl2018

Repository for competition https://www.kaggle.com/c/data-science-bowl-2018

Requirements

To run this code you need

  • pytorch installed with CUDA support
  • kaggle-cli installed and tuned

Download and unpack data

In the data directory run

''' $bash get_data.py '''

Run the bounding boxer training

Run this to start training the boxer part of the model: ''' cd train python train.py '''

You can do that until you are satisfied with the results. For the final tuning run

''' python train_stage2.py '''

Run the masker training

To start training the masker part of the model run this: ''' python train_masker.py '''

Here are some of the results: