# facebookresearch/dinov2

PyTorch code and models for the DINOv2 self-supervised learning method.

Repository: https://github.com/facebookresearch/dinov2
Canonical: https://ross.abutalabs.com/products/dinov2
Language: Jupyter Notebook
License: Apache-2.0
License Family: permissive
Last push: 2026-06-03T15:32:00+00:00

## Health v2 (maintenance only)
Score: 68/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 85, release rhythm 35, longevity 89
- inputs: {"age_days": 1253, "days_push": 91, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 13266, forks 1255 (observed 2026-08-28T04:11:02.921044+00:00)

## What it is
PyTorch implementation and pretrained models for DINOv2, a self-supervised vision transformer method from Meta AI that learns robust visual features without labels. The models produce high-quality visual embeddings usable directly with simple linear classifiers across many computer vision tasks.

## Use cases
- extract visual features from images without fine-tuning
- train a linear classifier on frozen image embeddings
- get backbone weights for image classification or segmentation
- compute image embeddings for retrieval or clustering
- apply self-supervised vision transformers to medical imaging
- align image features with text using dino.txt inference

## When to choose
- you need strong general-purpose visual features without labeled data
- you want a frozen vision backbone for downstream tasks like classification, depth, or segmentation
- you are doing research on self-supervised learning or vision transformers

## When to avoid
- you need the newest model line - the successor DINOv3 is recommended by the authors
- you need a lightweight CPU-only model for production inference
- you need multimodal training rather than feature extraction

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, image-processing, computer-vision
- domain: machine-learning, deep-learning, computer-vision, image-processing
- platform: python
- tags: self-supervised-learning, vision-transformer, pretrained-models, pytorch, feature-extraction, foundation-model, gpu

## Member repositories
- facebookresearch/dinov2 (main) score 68

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:02.921044+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:13:23.811365+00:00, confidence not recorded.
  - readme: https://github.com/facebookresearch/dinov2 (fetched 2026-08-28T04:11:02.921044+00:00, sha 1dde817017e6)
- Data as of 2026-08-30T08:39:29.467469+00:00.