Note We release MiDaS v3.1 for monocular depth estimation, offering a variety of new models based on different encoder backbones. This release is motivated by the success of transformers in computer vision, with a large variety of pretrained vision transformers now available.

Intel/dpt-beit-large-512

Depth Estimation • Updated Mar 11 • 26.9k • 3

Note This model gives the highest quality, but is also the most heavy in terms of computation as mentioned in the paper.

Intel/dpt-swinv2-large-384

Depth Estimation • Updated Mar 26 • 95

Note This model has moderately less quality, but has a better speed-performance trade-off

Intel/dpt-swinv2-tiny-256

Depth Estimation • Updated Mar 26 • 551 • 5

Note This model is recommended for deployment on embedded devices