Government, Public Safety, Smart Cities
The first nation to meet UN SDG 11 with AI
A country cannot inspect itself. Saudi Arabia covers 450,000 kilometres of surveyed ground, thousands of municipalities, and a public safety mandate that runs from an open manhole to an exposed wire. Manual inspection was never going to reach the end of it.
Aug 11, 2020
3 min read
The scale problem no inspection workforce can solve
The Ministry of Municipal and Rural Affairs and Housing (MoMRAH) oversees urban planning, city enhancement and municipal services across the Kingdom of Saudi Arabia. Its public safety mandate is national. Its detection method was manual.- Coverage that could not scale: Extending surveillance to cover the entire country was not feasible with on-site inspection teams.
- Category breadth: More than 43 public infrastructure and public safety categories had to be detected consistently, with subjectivity minimised.
- Manual review at volume: Thousands of images and videos required human monitoring, delaying incident resolution.
- No automated triage: Without automated reporting and incident prioritisation, the most urgent hazards were not reliably the first ones addressed.
The solution: a computer vision engine built for national coverage
Under the Quality of Life programme within the Saudi Vision 2030 framework, MoMRAH deployed CamCom’s AI platform to auto-detect and classify public safety elements from images captured by citizens on mobile devices and from field surveys.- Context-aware AI: Deep learning analyses spatial relationships, object co-occurrence and topology variations, so a violation is judged in context rather than in isolation.
- Scalable processing: Optimised inference pipelines handle millions of images annually.
- GIS-integrated detection: Geospatial analysis filters detections against business rules and improves contextual accuracy.
- Advanced training pipeline: Out-of-distribution detection, self-supervised learning and active learning drive continuous model improvement.
- Adaptive computer vision: The models keep evolving, supporting new use cases and adapting to regulatory change.
- Automated classification and assignment: Incidents are classified, mapped to the relevant subclause, and assigned to the appropriate department.
- Quantity and severity: The system detects violation quantities and severity levels, enabling genuine prioritisation rather than first-in-first-out.
- Auto-generated incidents: Each incident is created with the details needed to act on it.
- Central command centre integration: A dashboard surfaces key data points, with automated mobile and email notifications alerting the responsible officers.
- Geo-tagged visualisation: Violations are mapped for location-based analysis, and historical analytics expose high-risk areas and violation trends.
The results
- First nation in the world to achieve full compliance with UN SDG 11 — sustainable cities and communities.
- 62% reduction in fatalities and casualties.
- 83% faster resolution of public safety incidents.
- 67% faster case processing — down from three weeks to under one week.
- 90% reduction in the on-site inspection workforce.
- 91% boost in operational efficiency and a 76% reduction in operational costs.
- 96% expansion in surveillance coverage for public safety.
- $20M in revenue generated through penalties enforced for public safety violations.
- 7M public safety incidents processed every month.