STM32N6570-DK- Constant False Detections with BlazeFace on Edge AI Core v3.0

[Sradha]

 I am trying to deploy a custom quantized face detection model(blazeface_front_128_int8.tflite) on the STM32N6570-DK following the Deploy-your-Quantized-Model.md guide from the STM32N6-GettingStarted-FaceDetection repository. (STM32CubeIDE v2.0.0 and stedgeai v3.0).

Although the build and run are successful, the display shows constant false detections even with no face present.
Could you clarify if specific changes are required in "app_config.h ", and is there a recommended way to verify that the NPU is correctly accessing the "network_data.hex" weights from external flash?

[Julian E.]

Hi @Sradha,

 

How are you deploying the model? 

• Manually via the github
• Automatically with ST model zoo services

I would advise to test to deploy it with model zoo services and see if it helps

 

Maybe doing a validation on target could also help to understand if this is a firmware issue or an issue coming from the model.

I would suggest to use STM32 Cube AI Studio for that.

You can import input data and download the output of the inference on target.

 

Have a good day,

Julian