[Visionworks] Custom Node error with vx_scalar as output?

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX1
1

Hi guys,
I just created a custom node on openvx. I expected a parameter type vx_scalar as output. Because I need its change as input for the custom node right after.
But when I create a graph and create custom node in graph like:
vx_int32 nIters = 1;
vx_scalar nItersScalar = vxCreateScalar(context, VX_TYPE_INT32, &nIters);
vx_node custom_node = AbcCustomNode(graph, …, …, nItersScalar);

Everything fine until I call verifyGraph().
I dont know why the output error said that :
[NVX LOG] vxVerifyGraph: node[transform].parameter[3]: invalid scalar type INT32, expected INVALID.

You guys who has experience with openvx custom node or has faced with this problem can share with me some suggestions.? Many thanks!

My pleasure!
Asley.

2

Hi,

Please declaire the input / output parameter via vxAddParameterToKernel().

You can find a simple example here:
https://devtalk.nvidia.com/default/topic/1001961/jetson-tx1/visionworks-kernel-of-custom-node/post/5120722/#5120722

Custom node tutorial can be found here:

[i]>> VisionWorks Toolkit Reference Documentation

VisionWorks API
Tutorials
User Custom Node Tutorial[/i]

Thanks.

3

Ya, thanks @AastaLL
I forgot set meta data for vx_scalar output in validate callback.
When I add vxSetMetaFormatAttribute() for vx_scalar, everything works fine.

Thanks for your support!

Regards,
Asley.