(Error) Input tensor exceeds available data range

msee19018

Hello, I am trying to run a custom model for frame normalization. I did all the steps as explained in documentation for creating new custom model with pytorch. I created onnx file then simplified it and then created blob for that model.
When I load and run that blob on oak-d using NeuralNetwrok node it gives this error. I have attached photos of error and the code for you reference.

erik

Hello msee19018 ,
this looks like a difference between U8 and FP16. What datatype do you input to this custom model and what datatype does the custom model expect? If it expects U8, you would need to add -ip U8 to the compile tool, otherwise, you shouldn't specify it.
And as the documentation for this demo states;

This is just a demo, for normalization you should use OpenVINO’s model optimizer arguments
--mean_values and --scale_values.

Thanks, Erik