Error generating report after success Training 1D images (ECG Heart-Beats)

[mjrovai]

Error with generating report after success Training 1D images (ECG Heart-Beats).

I am working with some students to create a TinyML ECG Classifier.
We started from a real raw ECG dataset from Physionet and pre-processed data in order to generate fixed-length Heartbeats (100 points):

We upload the dataset as .CSV files and we created one impulse as Time series data (100ms of window size and increase) with no pre-process (Raw Data).

We defined our CNN model with good results after training:

But maybe because the automatic report generated after training was expecting a DNN model (usual with Time Series), we could not proceed with the Test/Deploy phase.

Downloading the Jupyter Notebook, we confirmed that the model was trained correctly’:

This is the project:
https://studio.edgeimpulse.com/studio/43912

Please let me know if I need to send more information.
Thanks
Marcelo

[louis]

Hello @mjrovai,

Can you try to add a reshape layer in between your Sequence model declaration and your conv1D layer:

I tried on my side and it should work

# model architecture
model = Sequential()
model.add(Reshape((int(input_length / 2), 2), input_shape=(input_length, )))
model.add(
    Conv1D(8,
           kernel_size=7,
           activation='relu',
           padding='same',
           input_shape=(features, depht)))

Regards,

Louis

[mjrovai]

Worked very well! Thanks a lot!
Marcelo