What we do
An open observation layer for the physical world
OpenSight™ is an open, public-interest initiative creating the common infrastructure for capturing, sharing, enhancing and understanding the physical world. Roads come first: a low-cost, open vehicle appliance turns the drives a council or contractor already makes into a shared, ongoing record of the network.
The same appliance, schema and principles apply beyond roads too, to footpaths, signs, drains, street furniture, vegetation and bridges. Roads remain the first domain and the proof case.
Core concept
The Observation
An Observation is a georeferenced, timestamped capture (image or video) with positioning quality, device, sequence, sources and provenance.
The differentiator
Change over time, not a snapshot
The valuable question is not "where are the potholes?" but "what has deteriorated on this road since the last inspection?"
A pothole appears. The system detects it, then recognises when the vehicle drives past again that the pothole remains and has changed. Once it is repaired, the next observation confirms the fix.
Design principles
Ten principles, adopted as our values
Open by default
Hardware agnostic
Offline first
Evidence first
Human-verifiable
Standards based
Privacy by design
Cloud optional
Vendor neutral
Commercially compatible
The data model
Three rules for the platform
OpenSight follows three rules for how the platform's data works, unchanged since the prototype was first built.
- An observation, not a recording, is the fundamental record: a georeferenced, timestamped capture with its positioning quality and provenance. Cameras are adapters; the record does not depend on any one of them.
- Mapillary is a partner and a distribution layer, not the foundation the platform is built on. Imagery contributed to it becomes open under CC BY-SA. The observation database and the evidence archive stay under the control of whoever captured them.
- Source imagery, derived observations and private organisational data are three different things, kept apart, each with its own licence intact.
Licensing and ownership
What open does not mean
Open source does not mean every dataset is free to redistribute. Data a council or contractor collects about its own network stays theirs to keep private, and nothing on the platform requires it to be shared. Imagery contributed to Mapillary is different: Mapillary requires it to be published under the CC BY-SA licence, and it stays under that licence once it is out. The observation database and the evidence archive on the appliance are not covered by that licence and remain under the control of whoever captured them. No single camera, cloud provider, AI model or GIS platform controls the ecosystem, so adopting OpenSight does not tie an organisation to one vendor's stack.
Mapillary's role
An integration, not a dependency
Mapillary is one imagery distribution and discovery layer with global coverage, APIs and contributor mechanisms: an integration, not a fundamental dependency. The platform must work with independently captured and stored imagery too. For now, though, it's a key part of how we share imagery.
About Mapillary
Mapillary (opens in a new tab) is a street-level imagery platform owned by Meta. Anyone can contribute imagery from a phone or a dashcam, and contributed imagery is published under the CC BY-SA 4.0 licence (opens in a new tab), so others can use it with attribution. It offers a public web viewer, contributor apps and a developer API (opens in a new tab), and OpenStreetMap editors use it too. You can read more at its help centre (opens in a new tab).
Imagery: Mapillary contributors, CC BY-SA. This is the credit line we use whenever their imagery appears.
The appliance
What OpenSight Edge is, in plain words
OpenSight Edge is the low-cost, open vehicle appliance at the front of the system: a small computer paired with a camera and a positioning receiver, fitted to an ordinary vehicle. It captures georeferenced imagery on every drive, keeps working without connectivity, and publishes what it captures once it can. It is built from open hardware designs so it isn't tied to a single camera, cloud or vendor.
In the interest of being clear
What we are not
OpenSight is not a finished product and does not claim to be. It is a prototype-stage, public-interest initiative built in the open, grounded in real vehicles, real roads and real engineering, not hype about what AI might one day do. Blurring of faces and number plates is a design requirement of the platform and is applied before imagery is published, for example via Mapillary. The appliance's local evidence archive is raw and stays with the vehicle owner under access control.