Inference hangs when using an motion sensor project

TinyML Board
1

Hello, I am working on a project which uses the motion sensor of the Syntiant TinyML board with edge-impulse. When I start the inference on the board (edge-impulse-run-impulse), the board doesn’t do anything and hangs after the initial message is printed on the terminal
Any idea what I am missing?
As a debug process, I also flashed the firmware prepared in this tutorial project - https://studio.edgeimpulse.com/public/102981/latest
but it was behaving in a similar way. I did get a couple of predictions (only 3) but there was a huge (like in minutes) random amount of delay b/w each prediction.

image
image856×458 91 KB

2

Hi @john_doe ,

Seems like it is not detecting any class.

May I know few things like

  • How did you collect the data set.
  • What are the 3 different classes for which you have trained the model.
  • What is the default LED Colour on the TinyML Board during inferencing.

As you said that circular motion example is working on the board (although only few detection)
it seems there is no hardware issue.

Thanks.

3

Hello,

  1. The dataset is collected using the same TinyML board via the edge-impulse studio.
  2. The different classes are Error, Normal_Motion, Z_No_Motion
    image
  3. Only the green LED is on which I believe is the power LED.
4

Hello @john_doe,

Can you provide bin file of the model so that I can test it .
We can setup a call also to resolve this together.

Thanks.