ESP32S3 classification random values

Help
#1

Hi everyone,

I’m experiencing an issue with my Edge Impulse model deployment on ESP32S3. The model compiles successfully, but it’s producing random/incorrect classification results when running on the device. Interestingly, the exact same model works perfectly when testing through the Edge Impulse web interface.

Technical details:

  • Device: ESP32S3
  • Type: Classification model
  • Build: Successful compilation
  • Issue: Inconsistent/random classification results on device
  • Web inference: Working correctly with expected results

Has anyone encountered similar behavior?

#2

Hi @eduard

Please try deploying this and uncommenting the following line - example-standalone-inferencing-espressif-esp32/main/CMakeLists.txt at 221c9c44cf73601e69070dc9df4bbef695752cbc · edgeimpulse/example-standalone-inferencing-espressif-esp32 · GitHub

Best

Eoin

#3

@eduard creating multiple forum threads and commenting in unrelated ones is not helpful.
What would be helpful is if you can follow the template.
The important thing for us to help you solve your problem is to be able to reproduce it on our own - or conclude we cannot reproduce it. So, we need the information about your environment and the code you used.
Also simply saying “Inconsistent/random classification results on device” does not tell us much about how does issue manifest itself.

ESP32S3 inferencing instable
#4