Jetson Nano ROS2 publisher and image processing

[muhammedsezer12]

Hi, I want to publish image via ros2 on jetson nano but I also would like to use it.
As far as I know, we can get img data from CSI cameras to publish directly.
But I also need to use it.

I cannot find on gitpages.

The question is: How can I get img data from jetson library and process image?
Also I would like to understand and know if I can process img while img is in jetson utils datatype (not numpy array)

I will use img in same node. I’ll publish image to use another devices.

thanks,

[dusty_nv]

Hi @muhammedsezer12, you can find ros2 package for CSI camera at https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_argus_camera

Also the one from jetson-inference / ros_deep_learning https://github.com/dusty-nv/ros_deep_learning#video_output-node

@muhammedsezer12

Also I would like to understand and know if I can process img while img is in jetson utils datatype (not numpy array)

You can use it with the functions from jetson.utils / jetson.inference, or see here for documentation about the memory structure: https://github.com/dusty-nv/jetson-inference/blob/master/docs/aux-image.md#image-capsules-in-python

[muhammedsezer12]

Hi, thanks.

Last question: is there any way to publish img from directly image capsule datatype? Without any conversation to opencv.

[dusty_nv]

The ros_deep_learning package doesn’t use OpenCV for the image conversion (it uses CUDA). But there will always be the step of serializing the image into the ROS message.

[muhammedsezer12]

Hi,

Real last question: how do you serialise img with your code?
How can i apply it on python?

[dusty_nv]

Hi @muhammedsezer12, you can find the code for converting CUDA image to ROS image message here:

https://github.com/dusty-nv/ros_deep_learning/blob/2b9b61288f7a93e2bbdb3ccc450bcf34a6d4cf04/src/image_converter.cpp#L159