A monocular depth estimation model using a large BEiT transformer backbone with DPT head, processing 512x512 inputs.