Note We introduce dense vision transformers, an architecture that leverages vision transformers in place of convolutional networks as a backbone for dense prediction tasks. We assemble tokens from various stages of the vision transformer into image-like representations at various resolutions and progressively combine them into full-resolution predictions using a convolutional decoder.

Intel/dpt-large

Depth Estimation • Updated Feb 24 • 162k • 175

Note This model leverages a Vision Transformer (ViT) backbone for monocular depth estimation.

Intel/dpt-hybrid-midas

Depth Estimation • Updated Feb 9 • 459k • 84

Note This model leverages a hybrid Vision Transformer (ViT-hybrid) backbone for monocular depth estimation.