Object detection project runs flawlessly on PC, but fails on STM32N6 hardware

Question/Issue:
[My Object detection project runs flawlessly when tested in a Web Browser using a web-cam etc, but the final port over to the STM32N6 dev-kit fails to detect anything when the Inference is running]

Project ID:
[848503]

Context/Use case:
[I have built an object detection project using 4 “toys” and an “unknown” category. The project detects the toys when tested using the web-browser option, but nothing is detected when the project is run on the STM32N6 dev-kit ]

Steps Taken:

  1. [Build project with image data]
  2. [Build project with more image data]
  3. [Step 3]

Expected Outcome:
[Objects are detected and labeled when the Impulse is run on the STM32N6 dev-kit]

Actual Outcome:
[Lots of “unknown” detections, but no object detection]

Reproducibility:

  • [*] Always
  • [ ] Sometimes
  • [ ] Rarely

Environment:

  • Platform: [STM32N6]

  • OS Version: [Ubuntu 22.04 and , Windows 11]

  • Edge Impulse Version (Firmware): [e.g., 1.2.3]

  • $ edge-impulse-run-impulse --raw
    Edge Impulse impulse runner v1.34.1
    [SER] Connecting to /dev/ttyACM0
    [SER] Connected to /dev/ttyACM0
    AT+INFO
    AT+INFO


  • Edge Impulse firmware *

Firmware build date : Dec 9 2025
Firmware build time : 10:25:04
ML model author : David
ML model name : stm32-n6-pokemon-Dec25
ML model ID : 848503
Model deploy version : 10
Edge Impulse version : v1.80.4
Used sensor : camera

Additional Information:
[Any other information that might be relevant]

HI @DJE666

I’ll check the project during the next days.

regards,
fv