I recently started my MSBA program at Cal Poly Pomona and had the opportunity to participate in a signature project. These projects are optional and the intent of the project is to gain new knowledge and skills. I have previously worked with DICOM images and have some experience with neural networks so I decided to make a few tutorials that our team could use to build a better understanding of preprocessing DICOM images and training neural networks.
Although we are not entering into the actual competition on Kaggle, this should be a great learning experience.
- Exploration.ipynb General tutorial and exploration of the DICOM files and dataset.
- Preprocess.ipynb Preprocessing tutorial.
- process.py Preprocessing file that uses multiprocessing to improve performance.
- error_log.csv Log of errors encountered while preprocessing.
- Training.ipynb Introduction into deep learning, Keras, and building/evaluating models.
- train_all.py Used for training several models on a p3.2xlarge EC2 instance in AWS.
- model_log_train1603151129.csv Log of the evaluation of each model trained in AWS.