Map Data Structure and Verification
Last updated
Last updated
The Hivemapper Network divides the world into trillions of small hexagonal tiles known as hexes. These hexes are the atomic unit of the map and are based on the H3 geospatial indexing system. Using small H3 cells ensures precise attribution of contributions and avoids ambiguity from partial or overlapping tiles.
When an approved Open Camera device (such as the Hivemapper Bee) contributes imagery, each contribution is evaluated at the hex level. Said differently, the atomic unit of contribution is a single image that captures useful map data within one or more hexes.
Examples of valid contributions include:
• Uploading fresh imagery for a hex that the network has flagged as stale or under-covered.
• Detecting map-relevant objects—such as lane boundaries, speed limit signs, toll prices, or traffic lights— within the boundaries of one or more hexes.
All contribution rewards and verifications are ultimately tied to these hex-based units, ensuring consistent, location-specific mapping across the globe.
Hexes are grouped into larger areas known as regions. The Hivemapper Network defines regions using standard administrative or statistical boundaries once contributors.
When you review the region pages on hivemapper.com, you will find useful data points about regions with active map contributors. These data points include the number of total and unique kilometers mapped, the number of contributors in the region, and the Region Progress score.
Verification is critical for maintaining the integrity and trustworthiness of the Hivemapper Network. Rather than focusing solely on human validation or hardware level validation, the system is designed to defend against fraudulent contributions such as AI-generated imagery, replay attacks, or other forms of manipulation.
Hivemapper uses a vision-based consensus mechanism operating at the hex level where each Bee device’s observations are cross-referenced with those from other devices. If multiple, independent devices report seeing similar objects (eg speed signs, turn restriction signs, traffic lights, toll prices, etc.) at the same location and time window, the system treats those observations as valid. This makes it extremely difficult for a single bad actor to inject fake or synthetic imagery without being detected.
To strengthen visual consensus and reduce the risk of coordinated manipulation, the Hivemapper Network incorporates device diversity weighting into its verification logic. Contributions from diverse devices—those with distinct hardware IDs, non-overlapping routes, and time-separated observations—are assigned greater weight in the consensus process.
Said differently, if two or more Bee (or ODC-compatible) devices are operating from the same vehicle or tightly correlated trajectories, their data is de-duplicated and rewards are not issued multiple times. These additional checks ensure that only authentic, independently captured data contributes to the network and earns rewards.
Each new Bee starts with low trust and gradually builds credibility by passing visual challenges and aligning with other devices’ detections. Early contributions might be throttled or require stronger consensus before they’re rewarded.
In low-traffic areas there may not be enough overlapping device coverage to establish consensus, leading to delays in validation or reduced rewards. However, this motivates good actors to deploy additional devices in low coverage areas thereby enhancing overall network coverage and freshness.
