Model training

As we mentioned earlier, we are using LSTM and CNN, particularly 1D CNN, for use case one. For use case two, we are using MobileNet v1. All of these DL implementations support transfer learning that does not require training from scratch to use them in IoT devices.

Use case one

We used an LSTM five-layered architecture for the ECG data classification, as shown in the following diagram:

We can train and test the model by running the LSTM_ECG.py file available in the chapter's code folder named use-case-1 subfolder:

python LSTM_ECG.py

For the ECG data in the remote patient-management system, we also tested and validated a CNN model. The following diagram presents the CNN architecture we used for the ECG...