Overview

DINOv2 was trained on 142 million curated images without captions. Its features work well across tasks with only a small task-specific head.

Highlights

  • Sizes from ViT-S (21M) to ViT-g (1.1B)
  • Self-supervised training
  • Strong frozen features
  • Apache 2.0 license

Typical uses

Image retrieval, depth estimation, segmentation and as a vision backbone.

Things to consider

  • Not a generative or chat model.

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.