🏛️OVER’s History & Evolution
The Starting Point
OVER launched in December 2020 with the vision of creating the largest decentralized platform for geo-localized Augmented Reality (AR) content. Publishing rights for these experiences are governed by OVRLand NFTs—spatial domains of 300 square meters mapped to real-world coordinates. These NFTs grant owners control over the AR content tied to their specific physical locations.
The project gained significant traction, quickly achieving key milestones:
Over 870,000 OVRLands sold
Distributed across 30,000+ unique wallets
More than 21,000 AR experiences published
OVRMaps: Building the World’s Largest 3D Map
However, this vision presented a significant technical challenge: how to precisely anchor AR content to a specific physical location? The solution to this problem became a central focus for OVER. Ultimately, it led to a pivotal shift that expanded the project's scope far beyond its initial AR focus and into the emerging field of Physical AI.
Anchoring AR content requires a device—whether a smartphone, AR glasses, or a VR headset—to know its exact position and orientation. While GPS is a common method, it lacks the necessary precision. It has an accuracy of 4-6 meters outdoors, is unreliable indoors (20-30 meter error), and provides no orientation data (i.e., the direction the device is facing).
The solution is a Visual Positioning System (VPS), a computer vision technique that determines a device's position and orientation in 3D space with centimeter-level accuracy. The catch? VPS requires a detailed 3D map of the location where it's being used. This makes high-quality 3D maps the most strategic asset in AR and the broader field of spatial computing.
To address this, OVER launched its Map2earn program. This DePIN (Decentralized Physical Infrastructure Network) framework incentivizes the OVER community with OVR tokens to create 3D maps, called OVRMaps, of key locations worldwide using standard smartphones and 360° cameras. The initiative has been a resounding success. To date, the community has mapped over 145,000 physical locations, creating the world's largest 3D map dataset, which includes:
72 million images
79 terabytes of raw data
The Pivot to Physical AI
We soon realized that the massive dataset generated by the Map2earn program had applications far beyond its original purpose. While essential for AR, these 3D maps are an invaluable resource for training a new generation of AI Vision Foundation Models, enabling machines and robots to understand and navigate the physical world.
These models are a cornerstone of Physical AI, a concept described by NVIDIA founder Jensen Huang as the next evolution of AI that allows machines to understand and interact with the physical world.
Similar to how the human brain processes visual information, these Vision Foundation Models — also called Large Geospatial Models (LGMs) — can interpret 2D images to generate a comprehensive 3D representation of a space, even inferring parts that are not directly visible. This process, known as machine perception, is inherently generative; it's about filling in the missing information to map 2D data into a 3D model. This allows machines to manipulate, move, measure, and stack objects, enabling applications in robotics, architecture, design, gaming and XR. The models achieve this by extrapolating from their training on the 3D structure of the world.
With its ever-expanding 3D map dataset from the Map2earn program - already orders of magnitude larger than comparable datasets -, OVER is uniquely positioned at the forefront of Physical AI. Our strategy involves both training our own state-of-the-art foundation models and licensing our unparalleled dataset to leading AI companies, protocols, and research institutions. This dual approach allows us to power the next generation of both centralized and decentralized AI systems, bridging the gap between cutting-edge research and real-world application.
OVER’s Core Value Propositions
This evolution has reshaped OVER's vision, which is now centered on four core value propositions:
Vision Foundation Model Training Developing and training proprietary AI models on our unique and growing dataset.
3D Map Data Licensing Selling access to our raw 3D map data for third-party AI training, robotics, and research.
VPS-Powered XR Publishing The original vision of AR/VR publishing, now enhanced with highly accurate content anchoring and expanded to B2B Whitelabel applications.
VPS API Services Offering our core Visual Positioning System technology as a service for developers and businesses to integrate into their own applications.
Last updated