ECG Hearthbeat Classification

Database

The notebooks are supposed to work on the PTB and MIT_BIH datasets.

Setup

The notebooks are supposed to run on Google Colab (it is possible to run it locally by changing the file paths and removing the drive mount for Google drive.

Packages

It may be required to install the PyWavelets library using for running the DBLSTM-WS and the ensemble models. The package may be installed by running the following command:

$ pip install PyWavelets

Train

In the first cells of each notebook you can find the following line:

train = False

This indiates that the model will be loaded from the specified .h5 file. and no training will be performed. To train the model simply change the previous cell to:

train = True

File paths

In the first cells of each notebook you can find 2 (in the case of transfer learning 3) paths:

data_path = "../input/"
model_path = "../../H5/<database name>/<model name>.h5"

The first path indicates the directory of the datasets, the second path indicates the file from which the model will be loaded (or saved if training is enabled). For the ensemble models, the model_path variable should point to the folder which contains the .h5 files:

model_path = "../../H5/<PTB || MIT>/"

Contributors

• @Matteo-Omenetti
• @ravifrancesco