Source code for training CNN on BCDR mammography images dataset
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data/BCDR-F03
yaml
.gitignore
README.md
classify.py
config.yaml
configure_cnn2.py
configure_cnn3.py
datasets.py
eval.py
fe_extraction.py
make_dataset.py
preprocessing.py
train_cnn.py
utils.py

README.md

cnn-bcdr

Source code for Representation learning for mammography mass lesion classification with convolutional neural networks (pdf).

Dependencies

This code is written in python. To use it you will need:

Usage

Data

With this paper we released the Breast Cancer Digital Repository F03 (BCDR-F03) dataset. You can get a copy from http://bcdr.inegi.up.pt/ (mirror at: https://bcdr.ceta-ciemat.es). Uncompress it under the data folder.

Preprocessing

  • Create hdf5 dataset:

    python make_dataset.py config.yaml
    
  • Build preprocessed version (GCN + LCN):

    python preprocessing.py config.yaml
    

Training

The hyperparameters to train the network are in the config.yaml file. Train the model:

python train_cnn.py config.yaml

Evaluation

Evaluate trained model:

python eval.py config.yaml