A 1-billion parameter monocular depth estimation model from Meta's Sapiens family for human-centric vision tasks.