Train models to find depth in a single photo.
Recent revisions are kept. Restoring one creates a new revision.
