Increasing the image publish rate rate in ROS

Hi,
Currently , I am working with Oak-d-pro on ROS2 and using camera.launch.py.

I have subscribed to topic /oak/rgb/image_raw. I see that camera is configured to work at 30fps, but I only see the topic publishing at 20 Hz.

average rate: 19.636

min: 0.003s max: 0.174s std dev: 0.03732s window: 2145

average rate: 19.629

min: 0.003s max: 0.174s std dev: 0.03732s window: 2165

Could you please help me in understanding this?

Thanks,
Chandra.

ChandrahasKasoju
Are you also using other streams - depth, left, right ,... these could slow down the pipeline.

Thanks,
Jaka

jakaskerl
Hi,
I just checked the publish rate after launching with another launch file "rgbd_pcl.launch.py".

The publish rate was as expected. It was indeed, the other launch file has launched NN which was slowing the pipeline.

I still have 2 questions:

1. I just need the rgb image values. Is there a way to only get the rgb images without stereo and pointcloud data?

2. I would like to increase to 60 fps. I have tried increasing the fps using dynamic reconfigure. But it did not work. How do I increase the fps?

Thank you for your support,
Chandra.

Hi,

You need to set pipeline type to RGB in camera config file. More information on parameter types can be found in the documentation
FPS is not a runtime parameter (runtime parameters have their names start with r, whereas "set" parameters start with i. That said, you can still change those parameters in runtime, you only need to restart the camera via ROS services. Other than that, changing camera config should suffice.
Few additional notes
FPS also depends on USB speed so make sure it's reported as SUPER/SUPER_PLUS
You can enable video encoding via i_low_bandwidth parameter, encoded frames are more lightweight so they can improve throughput between camera and PC, the driver then converts them automatically to same frames.
You can also use more advanced options to publish compressed ROS images (either CompressedImage for MJPEG or FFMPEGMsg for H264 encoding), more information on that can be found in the documentation
`rostopic
rostopic hz can be a little misleading with numbers when it comes to large messages such as images since it introduces additional overhead for few reasons such as needing do deserialize each message, being written in python and the fact that it isn't using ROS composition/IPC. If you integrate your own ROS nodes using those mechanisms then they should avoid those caveats.
You can try experimenting with different DDS implementations and/or tune current setup to improve performance, more information on that can be found in ROS docs