topic Re: Issue with Converting H5 Model for IMX95 NPU Deployment in i.MX Processors

Issue with Converting H5 Model for IMX95 NPU Deployment

varshilg — Thu, 01 May 2025 07:01:38 GMT

Dear NXP Support Team,

I am encountering an issue with deploying my custom model on the IMX95 NPU board. My model is currently in H5 format. I used the EIQ Portal to convert the H5 model to TFLite format, and then utilized the Neutron Converter for deployment. However, I am receiving an error indicating that the H5 format is not supported.

According to the user guide, H5 format should be supported as input. I have attached my model file for your review.

Could you please help me identify the cause of this issue and advise on the appropriate steps to resolve it?

Thank you for your assistance.

Re: Issue with Converting H5 Model for IMX95 NPU Deployment

Chavira — Thu, 01 May 2025 13:29:35 GMT

HI @varshilg!

Thank you for contacting NXP Support!

To convert your model h5 to tflite for TensorFlow Lite for Neutron you have to convert to tflite first .

With your labels in .txt format and other data according to your model you should be able to convert your model to a tflite format and then you can convert from tflite to tflite for Neutron

Best Regards!

Chavira

Re: Issue with Converting H5 Model for IMX95 NPU Deployment

varshilg — Thu, 01 May 2025 14:22:55 GMT

Hi @Chavira ,

I followed your steps and got below error.

Re: Issue with Converting H5 Model for IMX95 NPU Deployment

Chavira — Tue, 06 May 2025 14:41:16 GMT

HI @varshilg!

I am not sure if the script in eIQ has some error but you can try converting your model using the Python API and then convert the tflite model to Neutron model.