Hi community,
The micro_speech example for XIAO BLE sense using the link tflite-micro-arduino-examples/examples/micro_speech at main · lakshanthad/tflite-micro-arduino-examples · GitHub works only with TensorFlow version 1.15. My question is how to make it compatible with the TensorFlow version 2 generated tinyml model.
Kind regards,
Sandhya
I would like to know this ALSO ?
Hello Seeed?..Bueller , Anyone?
GL
PJ :-p
There are no plans for this at present. Your suggestion has been submitted upward. Thank you for your feedback!
I am currently trying the speech demo and got the version number error. I read that removing the dependency might work.
I commented out the tensorflow_version line with a “#” symbol.
#%tensorflow_version 1.x
import tensorflow as tf
I got past the version number issue and it is currently training the model (1hr into it). I don’t know if the model will work once I load it on the XIAO. Is there something else that will break after the training step?
Hi @pescatore,
The model trained using the current TensorFlow version doesn’t work on XIAO.
But hay there’s a Round Display?
Another “Halve of a Loaf” from SEEED ,I was disappointed by this discovery. It was all the rage in there promotional info on XIAO and touted as being Tensorflow capable…more of a Gimmick than Technological prowess.
Too bad They miss the Boat again. IMO
GL 
PJ
I am glad I checked before I bought anything. I suppose there is no way to go back to previous TF versions?
Not many solutions available for low cost simple word recognition.
I should check if the Arduino Nano 33 online example is for real.
Have also found this to be true two years later. I have succesfully trained 10 commands (0-9), but when I run on NRF52840 sense, get this message:
09:09:34.009 → Requested feature_data_ size 536871936 doesn’t match 1960
09:09:34.009 → Feature generation failed
The supplied library model.cc also only recognizes the “no” command… ignores “YES”…
any other microspeech tinkerers out there?