UniDepthV2 ViT-L/14

UniDepthV2 ViT-L/14 is a large-scale monocular metric depth estimation model that predicts real-world depth from a single RGB image without requiring camera intrinsics at inference time. It is the highest-capacity checkpoint in the UniDepthV2 family, offering metric 3D scene estimates with per-pixel confidence outputs across diverse camera types and scenes. Source: supabase.