OAK-SoM
A system-on-module vision platform for custom stereo depth and on-device AI in high-scale deployments.
656M
Total KM mapped.
36%
Global road coverage.
22M
Unique KM mapped.
OAK-SoM
OAK-SoM
656M
36%
22M
The Mission
Hivemapper’s smart, Luxonis-powered dashcams turn everyday vehicles into mapping contributors. They capture street-level imagery plus location and motion data to keep maps fresh for autonomous driving, logistics, GIS platforms, and utilities—without relying on expensive, centralized mapping fleets.
Vision Challenges
Mapping at global scale requires robust, low-cost edge perception:
- Capture high-quality RGB, stereo depth, and GNSS/IMU data while in motion.
- Avoid bulky LiDAR or RTK setups to keep hardware affordable.
- Withstand vibration and variable lighting in real-world automotive use.
- Process data on-device to avoid uploading raw video.
The Solution: Built with OAK
Hivemapper’s Bee and HDC‑S dashcams use a custom vision system built around the Luxonis OAK‑SoM. DepthAI processes synchronized stereo input, triangulates depth, and extracts map features directly on-device for efficient, privacy-preserving uploads.
Why Luxonis
Luxonis delivered a modular stereo + AI stack optimized for dashcam mapping.
Modular stereo + depth vision: The OAK SoM architecture made it easy to custom-build a stereo setup optimized for dash cam mapping without reinventing the wheel.
On-device AI and real-time compute: DepthAI handles image rectification, compression, and feature detection right on the camera, ensuring low latency performance with minimal connectivity.
Cost-effective scalability: With one core compute module used across all Hivemapper units, the platform scales globally without costly customization or reliance on server-based pipelines.
With Luxonis, Hivemapper can scale a global mapping network without the cost of traditional fleet-based systems.
Learn More
A system-on-module vision platform for custom stereo depth and on-device AI in high-scale deployments.
A decentralized mapping network powered by contributors and edge AI dashcams.