Setting rs.option.depth_units leads to shrunk space

[arthurgassner]

Required Info
Camera Model	D400
Firmware Version	05.13.00.50
Operating System & Version	Ubuntu 20.04
Kernel Version (Linux Only)	5.14.0-1052-oem
Platform	PC
SDK Version	2.45.0.3278
Language	Python
Segment	?

Issue Description

Dear Intel RealSense community,

I am using three Intel RealSense cameras in parallel (3xD430) and running them through Python.
All three cameras are connected to a USB hub (3.0), which is then connected to my computer through USB 3.0.

In Python, when setting up each device, I set the device's depth_units like so

pipeline = rs.pipeline()
profile = pipeline.start(rs_cfg)
profile.get_device().first_depth_sensor().set_option(rs.option.depth_units, 0.01)  # [m]

Consistently, the first time I run the image acquisition code, the generate depth image is shrunk on one of the images (compared to the others), i.e.

This is the case even all three cameras point roughly in the same direction

Restarting the image acquisition code fixes the issue.
Removing the profile.get_device().first_depth_sensor().set_option(rs.option.depth_units, 0.01) fixes it as well.

But as soon as I unplug-plug back the cameras, the issue appears the first time the image acquisition software is run.

Is this a known behavior? Am I interacting with the camera parameters wrongly?

Thank you for your time,
Arthur

[MartyG-RealSense]

Hi @arthurgassner As you are using the instruction first_depth_sensor, the SDK may be selecting one of the three cameras as the first depth sensor selected and then only applying the depth unit change to that first detected camera and not the other two, as they are not the first depth sensor detected.

An alternative approach may be to set up a separate pipeline for each camera with that camera's unique serial number, and configure an individual set of cfg instructions for each pipeline. An example of a Python script for doing this with two cameras that you could potentially adapt for three cameras can be found at #1735 (comment)

---

I also note that you are using kernel 5.14, which is not currently supported by the RealSense SDK. Unsupported kernels can work with the SDK but there may be unpredictable consequences in regards to stability. #10439 (comment) has an unofficial test patch for adding support for kernel 5.13 and 5.15.

[MartyG-RealSense]

Hi @arthurgassner Do you require further assistance with this case, please? Thanks!

[MartyG-RealSense]

Case closed due to no further comments received.