Question/Issue:
I’m trying to run an impulse on the Nano 33 BLE sense board. I have created a model that works very well in the Edge Impulse platform and with live classification using the Nano board. However, when I try to deploy it to the Nano hardware to run fully on the edge the classification values never change no matter the input. This doesn’t seem to be an issue when I use a spectrogram instead of the MFE but I don’t get the same performance with the spectrogram.
I get this message when I run the firmware on the board via the serial monitor:
WARN: run_classifier_continuous, enable_maf is true, but performance calibration is not configured.
Previously we’d run a moving-average filter over your outputs in this case, but this is now disabled.
Go to ‘Performance calibration’ in your Edge Impulse project to configure post-processing parameters.
(You can enable this from ‘Dashboard’ if it’s not visible in your project)
Predictions (DSP: 40 ms., Classification: 14 ms., Anomaly: 0 ms.):
metal-cutting-alarm: 0.00391
no-alarm: 0.99609
Predictions (DSP: 40 ms., Classification: 14 ms., Anomaly: 0 ms.):
metal-cutting-alarm: 0.00391
no-alarm: 0.99609
Predictions (DSP: 39 ms., Classification: 14 ms., Anomaly: 0 ms.):
metal-cutting-alarm: 0.00391
no-alarm: 0.99609
Predictions (DSP: 40 ms., Classification: 14 ms., Anomaly: 0 ms.):
metal-cutting-alarm: 0.00391
no-alarm: 0.99609
Audio classification
-MFE block
-Classifier (alarm/no-alarm classes)
Project ID:
221011
Context/Use case:
Audio classification of events for alarm