Replacing my own dataset in micro-speech example!

[leanh1234]

Hi guys!

I am researching the method to replace my own model in https://github.com/espressif/esp-tflite ... cro_speech. I also found the script for training the model: https://colab.research.google.com/githu ... pD52Na6jRa.

I try to modify from "yes/no" into other words and apply the model (array model) into micro_speech example. That always return 0.25 score for each class.
How can I fix that? Is that posible to replace my dataset to customize the wake word!
I hope I can receive any solutions from you guys!