DINOv3 - Classification

DINOv3 is a state-of-the-art self-supervised vision transformer developed by Meta AI. It learns powerful visual representations without labels, making it excellent for transfer learning and classification tasks.

Overview

DINOv3 for classification leverages self-supervised learning to create robust feature representations. Key features include:

• Self-supervised learning - Trained without labels using advanced techniques
• Powerful representations - Excellent transfer learning capabilities
• Vision transformer backbone - Based on ViT architecture with improvements
• Multiple model sizes - From small to giant variants
• Strong generalization - Works well across diverse domains with minimal fine-tuning

License

Meta DINOv3 License

Custom License with Restrictions

DINOv3 is licensed under Meta's custom DINO Materials license agreement. Key terms:

✅ Permitted Uses:

• Use, reproduce, distribute, copy, create derivative works, and modify
• Research and commercial applications
• Non-exclusive, worldwide, royalty-free license
• You own derivative works you create

⚠️ Restrictions:

• Must comply with Trade Controls and Sanctions
• Cannot be used for ITAR activities, military/warfare, nuclear applications, espionage, or illegal weapons
• Must acknowledge DINO Materials in publications
• Cannot reverse engineer or decompile
• Must redistribute under same license terms

📋 Legal:

• Provided "AS IS" without warranties
• Governed by California law
• License terminates if you breach terms or sue Meta over IP

Full License: Meta DINOv3 License

Pre-trained Model IDs

DINOv3 models must be trained on Roboflow or uploaded as custom weights. There are no pre-trained public model IDs available for classification tasks.

Custom model ID format: project-url/version (e.g., my-project-abc123/2)

Supported Backends

Backend	Extras Required
torch	torch-cpu, torch-cu118, torch-cu124, torch-cu126, torch-cu128, torch-jp6-cu126

Roboflow Platform Compatibility

Feature	Supported
Training	✅ Train custom models on Roboflow
Upload Weights	✅ Upload pre-trained weights (guide)
Serverless API (v2)	✅ Deploy via hosted API
Workflows	✅ Use in Workflows via Classification block
Edge Deployment (Jetson)	✅ Deploy on NVIDIA Jetson devices
Self-Hosting	✅ Deploy with inference-models

Installation

Install with PyTorch extras:

• PyTorch: torch-cpu, torch-cu118, torch-cu124, torch-cu126, torch-cu128, torch-jp6-cu126

Usage Example

import cv2
from inference_models import AutoModel

# Load your custom model (requires Roboflow API key)
model = AutoModel.from_pretrained(
    "my-project-abc123/2",
    api_key="your_roboflow_api_key"
)
image = cv2.imread("path/to/image.jpg")

# Run inference
prediction = model(image)

# Get top prediction
top_class_id = prediction.class_id[0].item()
top_class = model.class_names[top_class_id]
confidence = prediction.confidence[0][top_class_id].item()

print(f"Class: {top_class}")
print(f"Confidence: {confidence:.2f}")

# Get all class confidences
for idx, class_name in enumerate(model.class_names):
    conf = prediction.confidence[0][idx].item()
    print(f"{class_name}: {conf:.3f}")

Model Variants

DINOv3 is available in multiple sizes for different performance requirements:

Variant	Parameters	Use Case
Small	~22M	Fast inference, edge deployment
Base	~86M	Balanced performance
Large	~304M	High accuracy applications
Giant	~1.1B	Maximum accuracy, research