Flow - 4x faster than typing
Mini toolsDINOv2
DINOv2 is a family of Vision Transformer models (ViT-S, ViT-B, ViT-L, ViT-g) trained via self-distillation on a curated 142 million image dataset, without labels or text supervision. The models output a class token and patch tokens, optionally with register tokens, serving as general-purpose visual features. These features perform strongly out of the box with simple linear classifiers or k-NN, supporting image classification, semantic segmentation, monocular depth estimation, and image retrieval via nearest neighbor search, without requiring task-specific fine-tuning.
Image
Overview
DINOv2 is a self-supervised Vision Transformer developed by Meta AI that learns general-purpose visual features from unlabeled images. It produces class and patch token embeddings usable without fine-tuning for image classification, semantic segmentation, depth estimation, and image retrieval.
About Meta Platforms
We're connecting people to what they care about, powering new, meaningful experiences, and advancing the state-of-the-art through open research and accessible tooling.
Industry:
Technology, Information and Media
Company Size:
78865
Location:
Menlo Park, California, US
Website:
ai.meta.com
Last updated: July 15, 2026
