Making the camera pose randomized around a dome or hemisphere above the mean positions of objects of interest

[mona.jalal]

I am interested in a Python script that can hover the camera in a hemisphere or dome above the mean 3D position of objects of interest so that not only it’s not also the height from above but also it would create more diverse camera poses.

To give you an idea, the above view camera in offline dataset generation in warehouse has many limited views and is not interesting to me since it is not so randomized.

Could you please guide me in this regard with a YouTube or written tutorial?

For example, as you see some of the camera pose distribution in the created datasets below are quite diverse.

2205.07058.pdf

[Robo_qq]

maybe you can use lookat

[ahaidu]

Using Replicator randomizers you can generate a list of sample poses of the camera and then use this as input for the randomizer function, using choice or sequence.

You can also generate the orientations yourself and include them as a list as well, or you can use the look_at argument to generate this internally.

Here is an example of sampling poses on a sphere/dome:

• https://www.cmu.edu/biolphys/deserno/pdf/sphere_equi.pdf

And an example of using a list to randomize the camera poses:

• Camera Examples — Omniverse Extensions documentation