hemalathakkk/1

Kaggle-MobileODT-Cervical-Cancer

Classification of cervix types for cervical cancer screening. The classification is made from images and uses convolutional neural networks.

All of the code necessary for implementing the exploratory data analysis, benchmark models, preprocessing, and neural networks is in the folders modules and workflow_classes. For actually running these python scripts following a logical work flow, from beginning to end, the notebook FullWorkflow.ipynb is used.

Setting up necessary data

In order to run the main notebook FullWorkflow.ipynb, it is necessary to first acquire the training data. This is done from the Kaggle competition website: https://www.kaggle.com/c/intel-mobileodt-cervical-cancer-screening/data . It is only necessary to download the files test.7z and train.7z.

This data should be placed in a folder called Data which lies in the same directory as the jupyter notebook: the training data should be in the folder ./Data/train and the testing data should be in a folder ./Data/test. The train folder should have three subfolders: ./Data/train/Type_1, ./Data/train/Type_2 and ./Data/train/Type_3, each folder containing a lot of images.

Required python packages

The code has been written in the Anaconda distribution of Python 2.7. Additionally, the packages tqdm and TensorFlow need to be installed; the installation instructions are found here and here.