Meta Launched DINO to Green British Cities with Cost Savings

Meta Launched DINO for Urban Greening in British Cities with Cost Savings

Meta presented the application of its DINOv2 computer vision model in a UK urban greening project. The model helps British authorities analyze territories, identify suitable zones for new parks, and reduce costs associated with traditional land surveying. This is the first major case where self-supervised AI transitions from research laboratories into the hands of government structures.

How DINOv2 Works with Territory Analysis

DINOv2 is a self-supervised vision transformer developed by Meta based on the Vision Transformer architecture. The model was trained on 1.4 billion images and can analyze new data without manual labeling. In the context of urban greening, it studies satellite and aerial imagery, identifies soil types, existing vegetation, roads, buildings, water bodies, and territorial accessibility for development. The key advantage: the model operates in zero-shot mode—it requires no special preparation for a specific urban greening task. The British government can upload fresh satellite images, and DINO immediately provides analysis of suitable zones. This dramatically lowers the barrier to entry for government structures that previously had to hire expensive consultants and wait months for results.

Application in British Cities and Real Cost Savings

The UK government uses DINO to accelerate urban greening programs and increase residents' access to parks and open spaces. Previously, such analysis required weeks of manual work by expert surveyors and territorial planning consultants. Now the work cycle is shortened dramatically:

• Satellite imagery is processed in hours instead of weeks of research
• Costs for field surveys and expensive specialist visits are reduced
• Decision-making on expanding green zones happens orders of magnitude faster
• Budget is redirected from diagnostics to actual greening and improvement
• Fairness improves: all city neighborhoods receive the same objective analysis

The project is especially valuable for English and Scottish cities, where historically there has been a low percentage of park space per capita. According to British planning standards, each residential district should have access to a park within walking distance (typically 300-400 meters), but today this standard is far from being met in all neighborhoods.

Model Openness and Scaling Beyond the UK

Meta released DINOv2 under the open Apache 2.0 license. This means other governments, municipalities, and environmental organizations can adapt the model to their territories without paying Meta. The British case serves as proof of concept—demonstrating that AI can solve not only corporate tasks but also socially significant environmental problems.

"DINOv2 opens the possibility for Global South nations to analyze urban greening without expensive consultations,"

Meta emphasizes.

The project has already attracted the attention of environmental organizations and municipalities in other countries that are preparing similar pilots and planning to apply the model to forest and water resource analysis.

What This Means for Climate Policy

AI is transitioning from laboratories into government structures not as an experimental tool but as a practical way to save taxpayer money and accelerate climate programs. For climate policy, this means that routine territory surveys will become cheaper and faster, and government budgets will be able to focus on actual greening rather than long and expensive diagnostic work. The model also demonstrates that Meta's open-source approach to AI may be more effective for society than closed commercial systems.

Meta is recognized as an extremist organization and is banned in Russia.