Start a project
AI intelligence system

RoadCondition

From road condition to maintenance priority.

A survey system that converts vehicle-mounted road imagery into per-segment condition scores, deterioration trends and a maintenance programme ranked by cost of deferral rather than by complaint volume.

Government Public Infrastructure Transportation Computer Vision GIS
Sector
Government & Public Infrastructure
Year
2024
Duration
8 months
Team
6 people
Platforms
Vehicle capture · Web GIS
Status
In production
RoadCondition — visual identity for the case study
4,000km Network covered Continuous, not sampled
Per-segment Scoring granularity Matched to maintenance units
Cost of deferral Ranking basis Not complaint volume
Opportunistic Capture model Vehicles already on the network

01 — The problem

The authority maintained four thousand kilometres of road on a maintenance programme driven largely by public complaints. Roads in visible, vocal areas were resurfaced while quieter segments deteriorated past the point where surface treatment would have been enough.

Formal condition surveys existed but were manual, expensive and infrequent. By the time a survey was processed, the programme it informed was already a year behind the road.

02 — What we did

We made survey capture cheap enough to repeat. Cameras mounted on vehicles already driving the network — inspection vans, gritters, refuse trucks — produce continuous imagery with no dedicated survey run.

The scoring model was tuned to the decision, not the defect. Individual cracks matter far less than the segment-level trajectory, so the system reports condition per maintainable segment and ranks by projected cost of deferral, which is the number an engineer builds a programme from.

The cheapest repair is the one done before the segment needs rebuilding.
Interface

What it looks like in use

Two of the screens that carry the most weight in daily use, rebuilt here from the production design system.

roadcondition.aurezalabs.com/programme
Maintenance programme · FY264,012 KM
A-road seg 118 · surface window closingDEFER +$412k
B-road seg 44 · steady declineDEFER +$88k
Urban seg 902 · stableDEFER +$11k
62% of budget allocated

Ranked by cost of deferralSegments are ordered by what waiting will cost, which is the number an engineer actually builds a programme from.

roadcondition.aurezalabs.com/coverage
Network coverageOPPORTUNISTIC
Surveyed under 30 days2,880 km
Surveyed 30–90 days901 km
No recent pass231 km
Weekly km captured

Survey without survey runsCameras on gritters and refuse trucks provide continuous coverage, so the network is measured rather than sampled.

Capabilities

What the system does

Grouped by the job each set of capabilities exists to do, rather than by which team built it.

Capture

01 · Survey without survey runs

Vehicle-mounted ingestion

Imagery from cameras on operational vehicles with GPS and IMU logged alongside every frame.

Automatic segment matching

Frames matched to the authority's maintainable segment definitions rather than to arbitrary distances.

Coverage tracking

Network coverage and recency mapped so gaps are visible and routable.

Condition-aware quality control

Poor light, rain and speed effects detected and excluded rather than scored badly.

Assessment

02 · What the road is doing

Crack and pothole detection

Surface defect detection and classification aligned to the national condition standard.

Deterioration scoring

Segment-level condition scores computed from defect density, type and extent.

Historical comparison

Scores tracked across survey passes so trajectory, not just current state, is known.

Prediction

Projected condition at horizon, with the point at which treatment options narrow made explicit.

Programme planning

03 · The output that matters

GIS integration

Scores and defects published into the authority's existing GIS rather than a separate map nobody opens.

Repair prioritisation

Segments ranked by projected cost of deferral, balancing current condition against deterioration rate.

Treatment matching

Appropriate intervention suggested per segment, since a surface dressing and a rebuild are not interchangeable.

Budget scenarios

Programme outcomes modelled against different budget envelopes to support funding cases.

Architecture

How it is put together

Layer by layer, with the reason each one exists — because the reason is usually the interesting part.

01Capture kit

Low-cost vehicle units logging imagery with synchronised GPS and IMU, uploading opportunistically over cellular.

  • Vehicle cameras
  • GPS/IMU
  • Cellular upload
02Map matching

Trajectory snapping to the road network and association of frames with maintainable segments.

  • Map matching
  • PostGIS
03Detection models

Surface defect detection aligned to the national standard, robust to varying capture conditions.

  • PyTorch
  • Detection
  • Condition robustness
04Scoring and prediction

Segment score aggregation with deterioration modelling across survey history.

  • Scoring model
  • Trend projection
05Planning console

GIS-integrated workspace with prioritised programme, treatment matching and budget scenarios.

  • React
  • MapLibre
  • GIS export

Technology

Vision

  • PyTorch
  • Detection models
  • Image quality gating

Geospatial

  • PostGIS
  • Map matching
  • MapLibre
  • GDAL

Backend

  • Python
  • FastAPI
  • PostgreSQL
  • Celery

Platform

  • S3
  • Kubernetes
  • GPU batch
  • Airflow
Outcome

What changed

Measured against how the operation ran before, not against a benchmark chosen after the fact.

  • The programme stopped following complaints. Prioritisation is driven by condition trajectory and cost of deferral.

  • Survey became continuous. Vehicles already on the network provide coverage without dedicated runs.

  • Deterioration is visible before it is expensive. Segments are treated while surface options still exist.

  • Funding cases got evidence. Budget scenarios model programme outcomes rather than asserting need.

How it ran

  1. Months 1-2

    Standards alignment

    Mapped model outputs onto the national condition standard and the authority segment definitions.

  2. Months 3-4

    Capture and matching

    Vehicle kits, upload pipeline and trajectory-to-segment map matching.

  3. Months 5-6

    Detection and scoring

    Defect models with condition robustness and segment score aggregation.

  4. Months 7-8

    Planning tools

    Deterioration modelling, prioritisation, treatment matching and budget scenarios.

Related work

From the same practice

All fifteen projects

Manufacturing & Production Quality

LineTrace

From defect detection to defect causality.

A multimodal system that correlates line video, machine telemetry, batch records and operator actions to explain why a defect happened, not merely that it did..

2025 · 10 months
Telecommunications & Network Operations

NetCause

Not alarm detection. Failure hypothesis reduction.

A correlation engine that turns an alarm storm into a short ranked list of probable root causes by reasoning over network topology and fault propagation rather than alarm counts..

2025 · 8 months
Oil & Gas · Asset Integrity

PipeIntegrity

From a single inspection to longitudinal asset integrity.

A vision system that detects and localises pipeline deterioration, then tracks each defect across inspection cycles so maintenance is prioritised by progression rather than by appearance..

2024 · 9 months
Something similar?

Running into the same problem RoadCondition solved?

Tell us what you run. We will reply within two business days with what we would build for your situation — and, just as usefully, what we would leave out.

Start a conversation More work