Camera Focal Length and camera calibration in Sionna

[Leeyyhh]

Hi Everyone,
I’m currently working with the Camera class in Sionna.
I’d like to understand if Sionna’s Camera class uses traditional camera configurations to generate the images. My project requires using focal length and image resolution (instrinsic matrix)to accurately deduce object locations within the scene. But I couldn’t find explicit information on these parameters in the documentation.
Thanks a lot for your help!!

[merlinND]

Hello @Leeyyhh,

Sionna uses the rendering framework Mitsuba under-the-hood, and so renderings use a perspective camera that has a 3D pose (extrinsincs) as well as intrinsics.
Documentation: https://mitsuba.readthedocs.io/en/stable/src/generated/plugins_sensors.html#perspective-pinhole-camera-perspective

That being said, I'm not sure I understand the goal.
If you can render the Sionna scene, then the location of all objects in that scene are already known, you wouldn't need to go through the camera to deduce their location?

[merlinND]

Interesting! Then I would definitely recommend having a look at the Mitsuba documentation; if the goal is to simulate a real detection process you could improve the realism of the rendered scenes.

[Leeyyhh]

Hello, I was wondering if it might be possible to adjust the default camera's FoV in the sionna?

[merlinND]

Hello @Leeyyhh,

Yes, have a look at the function signature of scene.render():

sionna/sionna/rt/scene.py

Lines 1530 to 1533 in 0aeca34

	def render(self, camera, paths=None, show_paths=True, show_devices=True,
	coverage_map=None, cm_tx=None, cm_db_scale=True,
	cm_vmin=None, cm_vmax=None, cm_metric="path_gain", cm_show_color_bar=True,
	num_samples=512, resolution=(655, 500), fov=45):

The fov argument controls the FoV in degrees.

[Leeyyhh]

Thanks for the timely assistance! It really solved my problem and made everything much clearer. :)

[merlinND]

Glad it helped!