Closed lukicdarkoo closed 4 years ago
I agree the second solution is the best. It should however be clearly documented that only one topic can be used at a time and possibly display a warning or an error if the user tries to use two topics simultaneously.
Just for the note, a new node for the raspberrypi camera was released (ROS1), it would be nice to see if we are close to it (topics names, open parameters, etc): https://discourse.ros.org/t/raspberry-pi-camera-node/1388
For example, it would be interesting to publish the camera_info
topic.
Just for the note, a new node for the raspberrypi camera was released (ROS1), it would be nice to see if we are close to it (topics names, open parameters, etc): https://discourse.ros.org/t/raspberry-pi-camera-node/1388
Yes, raspicam_node
is available for ROS2 as well (https://github.com/christianrauch/raspicam2_node), we have already considered it. But, the camera on e-puck2 doesn't use Raspberry Pi Camera (OV760 instead) and it is not connected through standard CSI port (but through onboard MCU). All this makes raspicam_node
unusable for our use case. We also confirmed this with GCtronic.
For example, it would be interesting to publish the camera_info topic.
Sure, I will add an issue or implement it with the next camera-related PR. It is already implemented in webots_ros2_epuck2
(https://github.com/cyberbotics/webots_ros2/blob/64d48a40b6b687f7759a1ec4ec45c6cbd703d252/webots_ros2_epuck2/webots_ros2_epuck2/driver.py#L372)
Yes, raspicam_node is available for ROS2 as well (https://github.com/christianrauch/raspicam2_node), we have already considered it. But, the camera on e-puck2 doesn't use Raspberry Pi Camera (OV760 instead) and it is not connected through standard CSI port (but through onboard MCU). All this makes raspicam_node unusable for our use case. We also confirmed this with GCtronic.
Yes of course I am not saying we should use it, but just to be 'compatible' with it as much as possible.
Describe the Bug Topic
image_raw/compressed
publishes compressed image, however, there is a problem withimage_raw
. In order to be published, the image has to be converted from YUV to RGB and optionally resized. Unfortunately, the GPU block for image conversion is SISO (Single Input Single Output) and therefore, GPU cannot facilitateimage_raw/compressed
andimage_raw
simultaneously. Currently, it tries to use CPU for the conversion, but this approach is CPU intensive and also, OpenCV fails to perform the conversion.There are two possible solutions:
image_raw/compressed
orimage_raw
).I would prefer the second option since it is unlikely both topics will be used simultaneously. Furthermore, using the two topics simultaneously can cause very high CPU usage.
Steps to Reproduce Subscribe to
image_raw
topicExpected behavior User should be able to receive an image from
image_raw
topic.