Error's related to deployement of converted nef model in kneron hardware for inference
Hello Kneron Team,
I have made one vgg-16 model and converted that successfully to nef using the docker. Now I want to deploy that nef file to KL720. We have wrote the script for deployment of the model in the hardware and passing image and getting inference results from that. Till the point of deployment of the nef model, we are successful. Then we read image using cv2 till that we are successful. Now we have to do some form of preprocessing on that image and pass it to the model, at this step we got stuck and got some error.
I am attaching the zip file containing nef along with the python script we used and the image along with that, the screen shot of the error we got. We most likely did mistake in the preprocessing, and making of generic_inference_input_descriptor.
Please help me out in proceeding with the deployment of the model and getting inference from the image. We are using kneron PLUS.
Comments
Hi Haresh,
The error you got is "ApiReturnCode.KP_ERROR_INVALID_PARAM_12," which means you did not fill out kp.GenericInputNodeImage correctly.
Please refer to the right part of the code (KL720DemoGenericImageInference.py)
Inside GenericInputNodeImage are the parameters for preprocessing, so you can take the below code out.
The image dimension you're using is 399 x 162, with the width (399) as an odd number. This would give you an error, so it's recommended to adjust the image so that the width is an even number.