Library/dinov3Forked

facebookresearch/dinov3

dinov3

Reference PyTorch implementation and models for DINOv3

View on GitHub↗Upstream facebookresearch/dinov3↗

Builder

Meta Research

facebookresearch • ai-lab

Stars

10,005

Using upstream star count

Forks

800

Using upstream fork count

Open Issues

Activity Score

0/100

0 commits in 30d

Created

Aug 7, 2025

Project creation date

README Summary

DINOv3 is a self-supervised learning method for training vision transformers that produces versatile visual features without using labeled data. This repository provides the official PyTorch implementation and pre-trained models from Facebook Research. The method builds upon the original DINO approach with improvements for better feature quality and downstream task performance.

AI Dev Skills

Unmapped

Self-Supervised LearningVision TransformersContrastive LearningFeature ExtractionTransfer LearningPyTorch Model ImplementationRepresentation LearningDeep Learning Architecture DesignComputer Vision FundamentalsTransformer ArchitectureKnowledge DistillationComputer VisionModel Training and OptimizationPyTorch Deep Learning Framework

Taxonomy

AI Trends

Self-Supervised Learning Foundation Models Transformer Architecture Multimodal Vision Systems Vision Transformers Multimodal Reasoning

Deployment Context

Self-hosted On-premise Cloud API Cloud

Modalities

Image Vision

Skill Areas

Self-Supervised Learning Vision Transformers Contrastive Learning Feature Extraction Transfer Learning PyTorch Model Implementation Representation Learning Deep Learning Architecture Design Computer Vision Fundamentals Transformer Architecture Knowledge Distillation Computer Vision Model Training and Optimization PyTorch Deep Learning Framework

Use Cases

Unsupervised image representation learning Transfer learning for downstream vision tasks Image similarity and retrieval Feature extraction for classification Foundation model pretraining Unsupervised Visual Feature Learning Image Classification Transfer Learning Object Detection Feature Backbone Semantic Segmentation Initialization Image Retrieval and Similarity Downstream Vision Task Fine-tuning

Recent Activity

Updated 1 months ago

7 Days

30 Days

90 Days

Quality

research

Quality: high
Maturity: research

PM Skills

Developer Platform

Languages

Jupyter Notebook100.0%

Timeline

Project created: Aug 7, 2025
Forked: Mar 26, 2026
Your last push: 1 months ago
Upstream last push: 14 days ago
Tracked since: Mar 11, 2026

Similar Repos

pgvector cosine similarity · $0

Loading…