Rename to raw_image_pipeline

[mmattamala]

FYI: @YifuTao @rowanborder @nivedchebrolu @mauricefallon @ckassab @JonasFrey96 @plibera

This introduces important changes, as all the names that were image_proc_cuda* are changed to raw_image_pipeline* as it reflects better the fact that the code works independently of the hardware (CPU or GPU)

I also propose to change the repo name to raw_image_pipeline as it is Alphasense-independent and could be useful for other devices.

If we are all fine with this:

• I'll merge and we should not change the names anymore.
• We can start propagating these changes to the other repos in DRS and RSL.
• We should open source this repo as originally intended

[mmattamala]

@JonasFrey96 do you have any further comments on this? Otherwise I'll merge and will propagate the other changes to anymal_rsl

[JonasFrey96]

I will review this in the morning. Sorry for the delay.

[JonasFrey96]

Everything looks great and builds nicely.
Do we have instructions on how to tune the camera parameters and do the color calibration?
https://github.com/leggedrobotics/alphasense_rsl/tree/dev/rename-to-raw-image-pipeline/raw_image_pipeline_python/config

Minor issue: Can create a leggedrobotics forge of your pybind11_catkin.git such that everything stays under the same domain.

[mmattamala]

Awesome, thanks. I'll add more information on the different modules and how to do color calibration, @nivedchebrolu wrote a related issue #3

Regarding pybind, I'll check if there are any issues with the latest changes https://github.com/wxmerkt/pybind11_catkin
If everything works as expected, we can make a fork from that one instead (mine was just a temporary fix because the other one was outdated by the time I made the python wrapper)

[mmattamala]

Ok, some important changes:

• The README was improved to address Info on how to compute the color calibration matrix #3 @nivedchebrolu @JonasFrey96
• @lintong-zhang and I checked the namespaces of the output topics and now they are consistent for image and camera_info
  • /alphasense_driver_ros/cam4/debayered/camera_info
  • /alphasense_driver_ros/cam4/debayered/image
  • /alphasense_driver_ros/cam4/color/camera_info
  • /alphasense_driver_ros/cam4/color/image
  • /alphasense_driver_ros/cam4/color_rect/camera_info
  • /alphasense_driver_ros/cam4/color_rect/image

There are 3 namespaces now:

• debayered is the distorted image after correcting for the bayer pattern and flipping. This is good to record bags, as it can be compressed.
• color is the distorted, white-balanced image after white balance and other color corrections. This is nice for demos, but probably we need more work to find the right parameters. Not recommended for dataset recording.
• color_rect is the undistorted, white-balanced image.

Additionally, we can now indicate if the image is mono or color. If it's mono, it will publish to a different namespace instead mono_rect which is useful for visual odometry stuff.

• Lastly, I also added a flag to disable the image_transport plugins that are not used, such as theora or compressedDepth: link

[mauricefallon]

The Readme should have an image or video showing the different stages of processing. Could you add one?

[mmattamala]

@mauricefallon Yes, good call. I'll update the figure I used to have (see below) to reflect both the new updates to the pipeline as well as what's being published now

[rowanborder]

@mmattamala There's also a typo in the output_preffix parameter

[mmattamala]

Ok, merging this! New changes will be introduced as new PRs