CVPR002 Repurposing Diffusion-Based lmage Generators for Monocular Depth Estimation