Overview

Depth Anything V2, from researchers at HKU and TikTok, improved detail and robustness by training a teacher on synthetic images and distilling it on real images.

Highlights

  • Small (25M), Base and Large versions
  • Single-image depth maps
  • Fine detail on edges
  • Small model under Apache 2.0

Typical uses

Depth effects, 3D reconstruction, robotics and AR.

Things to consider

  • Base and Large weights are non-commercial.

Specifications are taken from the model card on Hugging Face and the provider’s announcement. Benchmark results depend on evaluation settings; see how AI benchmarks work.