Use the pyrealsense2 library to interact with the two Intel Realsense depth cameras on Stretch 3. The library can configure camera parameters, collect imagery, calculate point clouds, and much more. Start by creating a "pipeline", which is an interface for streaming data from the camera.
import pyrealsense2 as rs
cam = rs.pipeline()
cam.start()
cam.stop()start() returns a "profile", which returns information about the Realsense camera.
profile = cam.start()
print(profile.get_device().get_info()) # "D435if"
We pass a "config" to start() to choose which camera we want to stream from, what resolution the imagery should be, what frames per second the data should arrive at, and more.
config = rs.config()
config.enable_device(d405_info['serial_number'])
width, height, fps = 640, 480, 15
config.enable_stream(rs.stream.depth, width, height, rs.format.z16, fps)
config.enable_stream(rs.stream.color, width, height, rs.format.bgr8, fps)
profile = cam.start(config)
https://github.com/IntelRealSense/librealsense/tree/master/wrappers/python#python-wrapper
https://intelrealsense.github.io/librealsense/python_docs/_generated/pyrealsense2.html