Nvidia Visionworks

Hi, I wanted to know whether there is any API related or equivalent to opencv findContours?

Hi,
There is no findContours related API now. Here can find supported primitives of visionworks:

But you can refer to nvx_sample_opencv_npp_interop, which shows the interoperability of VisionWorks with other libraries, such as OpenCV.
nvx_sample_opencv_npp_interop can be found in:
$ ~/VisionWorks-1.5-Samples/samples/opencv_npp_interop/main_opencv_npp_interop.cpp

Hello, Thank you AastaLLL. I couldn’t find in that sample code. I wanted to know if there is any other API that is related to this.

You can copy sample to home by

/usr/share/visionworks/sources/install-samples.sh ~/
cd ~/VisionWorks-<visworks_ver>-Samples/
make -j4

Information can be found in:
https://developer.nvidia.com/embedded/dlc/visionworks-1-5-3-readme-linux

Hello ,
I couldn’t find equivalent API for findContours. Is there anything for boundingRect and ApproxPolyDP equivalent API ?

There is no findContours related API in Visionworks.

Possible alternative is to combine use opencv4tegra.
Please check nvx_sample_opencv_npp_interop sample which shows the interoperability of VisionWorks with opencv library.

Hello,
Thank you so much for the answer.

Anything for boundingRect API? I am writing a code with Visionworks APIs alone.

No boundingRect equivalent API in Visionworks.

Here can find the supported function of Visionworks:
https://developer.nvidia.com/embedded/visionworks
-> VisionWorks includes the following primitives

Not sure this is the correct place to ask this but here goes.

I’m using the current version of Visionworks to do some basic statistics on the features found in an image.
No problems using vxMapArrayRange to get the features across into host memory.
But when I try to interrogate the vx_keypoint_t structure I get weird integer values for the

vx_int32 x; x coordinate
vx_int32 y; y corodinate

They are changing from frame to frame in an encouraging way but don’t seem to represent points in the image plane.

Before I embark on some random bit masking, can someone enlighten me as to how the x,y position integers are represented in the vx_int32?
I’ve had a dig around on the web but there’s not much out there at present.

Here’s hoping the answer is trivial.

Thanks for any help.

Martin

Hi Marin,

Thanks for your question and sorry for my late reply.

There is another relevant issue using vxMapArrayRange and vx_keypoint_t.
Could you check if this topic helps first?
https://devtalk.nvidia.com/default/topic/999386/visionworks-openvx-on-tx1-vxmaparrayrange

Dear all,

I followed instruction about Visionworks 1.6, but when try to run a program, it gives error:

VisionWorks library info:
	 VisionWorks version : 1.6.0
	 OpenVX Standard version : 1.1.0

Error: CUDA-capable GPU was not found.

How to compile visionworks demo programs with GPU? My system is Ubuntu 18.04, GTX1080, amd64, 32 GB RAM.

Thank you very much in advance.

Warmest Regards,
Suryadi