cancer-img-classifier

Preparing a model for this breast histopathology image dataset on kaggle. This model takes in image data and classifies regions as either:

• 0 for the absence of Invasive Ductal Carcinoma (IDC) in an image crop.
• 1 for the presence of IDC.

Processing data

The data is available for download in the link above. The script data_unzip.sh handles unzipping the files and storing them in a folder ./data, which the code in data_extraction.ipynb expects it to be. For the time being, a working small subset of this dataset is stored (via pickle) in dataset.npys.

Neural network

A working neural network is in classifier.ipynb. Currently, it is coded to make sure it works for training. While I'm away from my main working machine, I'll be improving this script on the kaggle servers. I'll post a link here soon.

Dependencies

Please refer to environment.yml (I still need to remove scipy from it though). If you want to be up and running, run conda env create -f environment.yml.