[gmiller123456]

From the abstract[1], the only input is selecting a few of the repeating elements, and drawing motion lines. Depth and background are detected automatically.

[1] https://pixl.cs.princeton.edu/pubs/Willett_2018_AMI/