keras/dinov3_vit_large_lvd1689m

Model Overview

Vision Transformer (ViT) and ConvNeXt models trained using the DINOv3 method.

Reference

• DINOv3: Learning Robust Visual Features without Supervision

DINOv3 offers a powerful, generalist visual backbone learned entirely from unlabeled images as described in DINOv3: Learning Robust Visual Features without Supervision.

Links

• [DINOv3 Quickstart Notebook] - coming soon
• [DINOv3 API Documentation] - coming soon
• [DINOv3 Beginner Guide] - coming soon
• KerasHub Model Publishing Guide

Installation

Keras and KerasHub can be installed with:

pip install -U -q keras-hub
pip install -U -q keras

Jax, TensorFlow, and Torch come preinstalled in Kaggle Notebooks. For instructions on installing them in another environment see the Keras Getting Started page.

Presets

The following model checkpoints are provided by the Keras team. Weights have been ported from: https://huggingface.co. Full code examples for each are available below.

Preset name	Parameters	Description
dinov3_vit_small_lvd1689m	21.6M	Vision Transformer (small-sized model) trained on LVD-1689M using DINOv3.
dinov3_vit_small_plus_lvd1689m	29M	Vision Transformer (small-plus-sized model) trained on LVD-1689M using DINOv3.
dinov3_vit_base_lvd1689m	86M	Vision Transformer (base-sized model) trained on LVD-1689M using DINOv3.
dinov3_vit_large_lvd1689m	300M	Vision Transformer (large-sized model) trained on LVD-1689M using DINOv3.
dinov3_vit_huge_plus_lvd1689m	840M	Vision Transformer (huge-plus-sized model) trained on LVD-1689M using DINOv3.
dinov3_vit_7b_lvd1689m	6.7B	Vision Transformer (7B-sized model) trained on LVD-1689M using DINOv3.
dinov3_vit_large_sat493m	300M	Vision Transformer (large-sized model) trained on SAT-493M using DINOv3.
dinov3_vit_7b_sat493m	6.7B	Vision Transformer (7B-sized model) trained on SAT-493M using DINOv3.

All the models weights are under DINO V3 license: https://ai.meta.com/resources/models-and-libraries/dinov3-license/