How to deploy ML model on I.MX93 using NXP EIQ

[MridulBajaj]

Hello NXP Community,

I am a beginner in the field of AI/ML and currently working on deploying a machine learning model on the i.MX93 evaluation board.

I have connected the i.MX93 board using two USB Type-C cables – one to the power port and the other to the debug port. Additionally, an HDMI cable is connected to a monitor to display the output. The board is powered on and appears to be functioning normally.

However, the device is not appearing in the "Remote Device" list on the development environment. I am unsure why the board is not being detected and would appreciate any guidance regarding this issue.

Could someone please help me understand:

• Why the i.MX93 might not be showing up as a remote device?

• How to troubleshoot this detection issue?

• What should be the next steps to successfully deploy an AI/ML model on the i.MX93 board?

Any assistance or suggestions would be greatly appreciated. 

[danielchen]

Hi

Can you add device?

[MridulBajaj]

Hi,

I’ve connected my i.MX93 board to my PC via Ethernet — I can ping it and SSH into it without any issues. So the network connection is working fine.

However, I’m unable to add the board as a remote device in the eIQ Toolkit because the eiq_agent seems to be missing. I searched the board and couldn’t find the agent or any related binaries like eiq-ml-tool. I also couldn’t find a download link for the agent on the NXP site or in the documentation.

Would really appreciate help on where to get eiq_agent for i.MX93 or how to get the device properly added.

Thanks!

[danielchen]

Hi  

From below eIQ inference with DeepViewRT page,  the supported device include imx8m / imx8mini / imx8 nano/ imx8mp / rt1064 /rt1170.    imx93 is not in the supported device list

nxp.com/design/design-center/software/eiq-ai-development-environment/eiq-inference-with-deepviewrt:EIQ-INFERENCE-DEEPVIEWRT

[kual_gettobyte]

So how can we deploye trained and converted AI model from eiQ Platform to I.mx93 MPU.

We have FRDM BOARD OF I.MX93, and we want to build Edge AI applications on it. 

We are using eiQ platform for training and developing AI Models.

How can one deploy  Edge AI models onto I.mx93 mpu using eiQ platform?

Or I.mx93 doesn't not support running Edge ai applications?

[danielchen]

HI @kual_gettobyte :

To deploy edge AI models onto imx93 mpu, I would suggest 

step1:  quantize the model to TFLite model using eIQ toolkit command line or open-source tools (for example,  TFLite converter,  onnx2tf,  ultralytics for yolo, etc).

step2：  Run the quantized TFLite model on imx93 NPU with ethosu delegate

More details can be seen in i.MX Machine Learning User guider and eIQ toolkit UG.

Regards

Daniel