Error using multiple classifiers in a project on extract code


We are trying to use two classifiers in an impulse project. One to classify sounds (Microphone) and another using the (Gyroscope). The target board is the Arduino BLE 33 sense. When we reach the final part to extract the Arduino code, we get the below error:
Can only generate code for a single Keras block
Is there a specific way to have a single project with two networks and get the final code for implementation?

Attached screenshot

Thank you

Hi @Ghattas,

At this time, only one Keras-based machine learning block (e.g. Classification) is supported. You can have a ML block + an anomaly block, but not 2+ Classifiers. Hope that helps!

Hello Shawn

Many thanks for the advise
Do both blocks requires the same amount of features? we have received the below error

I have a thread with the same question: Different processing blocks input to different models - #13 by jarnold1

Hi @Ghattas,

This is a known issue at the moment and being worked on. Try selecting both “spectrogram” and “spectral features” for both your Classification and Anomaly Detection blocks to see if that allows you to train.

Thank you! we will check the thread and the discussion

Thank you! we will try the proposed approach to see if it works

1 Like

I think the main issue is if you don’t want all the features as an input to the classifier. In the Anomaly Detection block, some features can be unchecked, but this is not an option for the NN Classification block.

@shawn_edgeimpulse What is the expected timeline for a resolution to this bug?