This code uses two main scripts to train the model and classify the data:
train_model.py
Train your model. Add your model code to thetrain_12ECG_model
function. It also performs all file input and output. Do not edit this script or we will be unable to evaluate your submission.driver.py
is the classifier which calls the output from yourtrain_model
script. It also performs all file input and output. Do not edit this script or we will be unable to evaluate your submission.
Check the code in these files for the input and output formats for the train_model
and driver
scripts.
To create and save your model, you should edit train_12ECG_classifier.py
script. Note that you should not change the input arguments of the train_12ECG_classifier
function or add output arguments. The needed models and parameters should be saved in a separated file. In the sample code, an additional script, get_12ECG_features.py
, is used to extract hand-crafted features.
To run your classifier, you should edit the run_12ECG_classifier.py
script, which takes a single recording as input and outputs the predicted classes and probabilities. Please, keep the formats of both outputs as they are shown in the example. You should not change the inputs and outputs of the run_12ECG_classifier
function.
You can run this classifier code by installing the requirements and running
python train_model.py training_data model
python driver.py model test_data test_outputs
where training_data
is a directory of training data files, model
is a directory of files for the model, test_data
is the directory of test data files, and test_outputs
is a directory of classifier outputs. The PhysioNet/CinC 2020 webpage provides a training database with data files and a description of the contents and structure of these files.
The driver.py
, get_12ECG_score.py
, and get_12ECG_features.py
scripts must be in the root path of your repository. If they are inside a folder, then the submission will be unsuccessful.
See the PhysioNet/CinC 2020 webpage for more details, including instructions for the other files in this repository.