Depth Pro

Depth Pro is a foundation model for zero-shot metric monocular depth estimation, designed for researchers and developers who need to generate high-resolution depth maps quickly. It synthesizes depth maps with unparalleled sharpness and high-frequency details without relying on camera intrinsics. The model is also fast, producing a 2.25-megapixel depth map in 0.3 seconds on a standard GPU.
Depth Pro works by utilizing an efficient multi-scale vision transformer for dense prediction, a training protocol that combines real and synthetic datasets, and dedicated evaluation metrics for boundary accuracy in estimated depth maps. It also includes state-of-the-art focal length estimation from a single image.
Researchers and developers in the field of computer vision, especially those working on applications requiring fast and accurate depth estimation, will get the most value from Depth Pro. It can be used for a variety of applications, including robotics, autonomous vehicles, and 3D reconstruction, where precise depth information is crucial.