I am designing a new building(architecture) indoor dynamic mapping solution. Hardware is based on array of intel realsense cameras & intel provided SLAM. I have strong feeling you may be right skillful shoulder to tap to provide me some much needed middleware software that I require to complete my desired business workflow as follows:
1/ Custom utility to read intel realsense ROS bag file pulling out L515 data & writing accurate point cloud in PCD format with intensity.
2/ Custom utility to read intel realsense ROS bag file pulling out D455 data & writing color point cloud in RCS format.
3/ Custom utility to read intel realsense ROS bag file pulling out D455 data & writing video in WebM format with lossless codec.
4/ Custom utility to read intel realsense ROS bag file pulling out T265 data & writing accurate track in GPX format.
FYI: My current plan is to connect 7cameras in array including: 4x L515 + 2x D455 + 1x T265. I believe I can run all at same time & use intel realsense viewer to save as one ROS bag file.
Please advise how you can assist my team with above four utilities. Then my team can run with those outputs & do our magic with reconstruction into panoramic tours & smart objects: walls, windows, doors, etc … for architectural & engineering design processes.
Look forward to hearing from you sooner than later.
I see that several dozen folks have seen this post but I have not gotten any replies yet unfortunately which I am surprised by as this should not be considered a complicated request for the right person. If someone can please at least make a recommendation for a team or individual that I could speak to about this … a simple introduction would be Greatly appreciated as well. Just to clarify this is to be a well paid contracting opportunity tbd.
“My current plan is to connect 7cameras in array including: 4x L515 + 2x D455 + 1x T265. I believe I can run all at same time & use intel realsense viewer to save as one ROS bag file.” <-- Have you actually tried this? RealSense cameras tend to use a lot of bandwidth and power. Unless maybe you’re using MIPI instead of USB.
Hello John. I have only tested 2 cameras(L515 + T265) together. There were no issues. Intel RealSense team wrote about running multiple L515s together. The new D455 has not shipped as of yet. Yes this potential USB limitation issue would be an interruption to my current plans so I will have to test this immediately upon receiving additional unit(s). Thank you for this heads-up. For now I am assuming this is a surmountable issue(perhaps using MIPI as you suggest or simply running additional CPU+GPU(s) perhaps) & therefore I believe it is still safe to proceed in parallel with pursuing my desire for post processing software development. I will reach out to PickNik.
Yeah, definitely worth experimenting with. I know the DARPA Sub-T Challenge team that I was briefly a part of gave up on USB with four D435s because of how unreliable it was (with a beefy CPU). I think they were switching to MIPI.
Just to be clear, the problem isn’t the processing power of the CPU or GPU, but the bandwidth and power available over USB (and maybe even PCI).
The software requirements you describe above, though, look totally doable.
Can I assume that those who have tried to connect four RealSense cameras in past couple years were using USB3.0 which has theoretical limit of data speed 5Gbit/s? I am planning on using USB3.1 which has theoretical limit of data speed 10Gbit/s. I am hopeful this slight increase will enable desired better results.
As a freelance software engineer with more of 10 years in clean software development and ROS experience, I could implement your four tasks. Please, see my CV and contact me by Linkedin if you are interested in my service.