If you have trouble with your audio accuracy today, that’s our bad! The default neural network architecture that we load has the wrong reshape operation. To fix this, hover over the reshape layer, and click
Then select the
Number of coefficients parameter in your MFCC block (probably 13).
Accuracy is now back to where it was before
We’ll push a hotfix out later today.
This has now been resolved!
I have another problem today: after flashing my arduino nano 33 sense with a voice recognition network, running the firmware I find on device the gesture recognition network
@gransasso Are you sure you’re flashing from the right folder?
yes, i always delete old folder
@gransasso, wow you’re completely right. There’s two firmware files in the deployment now. Delete
arduino-nano-33-ble-sense.bin from the folder and reflash, that should solve it.
Will get fix out asap.
Thank you, now it’s running correctly. I had noticed too many items in the folder but had not focused on individual files
@gransasso thanks for reporting back. We’re pushing a fix out in 30 min, so thanks for finding this!
edit: scratch that still not OK
September 24, 2020, 7:20pm
This issue has been resolved now!