# facebookresearch/dinov3

Reference PyTorch implementation and models for DINOv3

Repository: https://github.com/facebookresearch/dinov3
Canonical: https://ross.abutalabs.com/products/dinov3
Language: Jupyter Notebook
License: NOASSERTION
License Family: other
Last push: 2026-07-15T13:57:04+00:00

## Health v2 (maintenance only)
Score: 59/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 92, release rhythm 35, longevity 27
- inputs: {"age_days": 391, "days_push": 49, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 11249, forks 939 (observed 2026-08-28T04:10:46.542522+00:00)

## What it is
Reference PyTorch implementation and pretrained models for DINOv3, Meta's self-supervised vision transformer backbone family. It includes training, evaluation, and probing recipes for tasks like segmentation, depth estimation, and image classification.

## Use cases
- extract visual features from images with a pretrained vision transformer
- fine-tune a self-supervised vision backbone for classification
- run linear probing for semantic segmentation on ADE20K
- estimate monocular depth from images
- train vision models on satellite or microscopy imagery
- download DINOv3 weights for use with Hugging Face Transformers or timm

## When to choose
- you need a strong general-purpose vision backbone for downstream tasks
- you want reference PyTorch code for self-supervised vision training or distillation
- you work on remote sensing or scientific imagery and want metadata-guided adaptation recipes

## When to avoid
- you need a lightweight model for edge or mobile deployment
- you want a non-PyTorch framework like TensorFlow or JAX
- you need a permissively licensed model for commercial use without checking the custom license

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

## Member repositories
- facebookresearch/dinov3 (main) score 59

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:46.542522+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 2026-08-29T17:16:25.062978+00:00, confidence not recorded.
  - readme: https://github.com/facebookresearch/dinov3 (fetched 2026-08-28T04:10:46.542522+00:00, sha 2d708fbb8aec)
- Data as of 2026-08-30T08:39:29.467469+00:00.