Dhaka deploys AI traffic enforcement system

Dhaka Metropolitan Police has activated an AI-run traffic enforcement system that links existing traffic cameras to automated violation-detection software, Reuters reports. Per The Daily Star, the rollout in April placed 25 AI-based cameras at several major intersections, including Shahbagh, Banglamotor, Karwan Bazar, Bijoy Sarani, Jahangir Gate and Airport Road. The Daily Star and PressReader report the system flags five violation classes: red-light jumping, entering closed left lanes, lane violations, wrong-way driving and illegal parking. Reuters quoted traffic sergeant SM Nazim Uddin saying drivers have begun to comply more frequently since the cameras were introduced. A motorist, Hannan Rahman Jibon, told Reuters he received an automated text and a 2,000 taka fine after running a red light. Local outlets including Dhaka Tribune and PressReader reported 200 to 300 automatic cases within the first days of operation.

Reporting by Reuters attributes the deployment to software that analyzes feeds from existing CCTV cameras to identify offences and generate cases for enforcement. The Daily Star describes the system as the "AI-based Road Transport Act Violation Detection System" operated by the Dhaka Metropolitan Police and notes current coverage is limited to key corridors. An IEEE conference paper from 2025 describes earlier Bangladeshi ITS work that used YOLOv8 for vehicle detection and number-plate extraction, reporting number-plate detection near 95% and vehicle-type detection above 85% in highway test data; that paper represents academic work on related components rather than the operational police system.

Editorial analysis: City-scale automated enforcement deployments commonly reuse existing CCTV infrastructure and rely on computer vision models for object detection, plate OCR, and rule-based violation logic. These systems typically detect a limited set of infractions first, as reported in Dhaka, because precision requirements and legal evidentiary standards constrain scope. Industry deployments often pair rapid, visible enforcement with messaging to increase deterrence; Reuters and local outlets describe early public-facing fines and text notifications in Dhaka.

Editorial analysis: For practitioners, Dhaka's deployment is a practical example of AI applied to urban operations under high-density, low-speed traffic conditions. The city's cited 4.8 kilometre per hour average speed, reported by Reuters citing World Bank and BUET data, highlights an operational environment where even modest behavioral shifts can yield measurable throughput and safety gains. The hybrid evidence base of media reporting and academic projects underlines a common pattern: research prototypes and ITS papers, like the IEEE study, inform but do not fully describe live municipal systems.

Editorial analysis: Observers should track three categories of indicators. First, measurement data, including changes in violation counts, crash rates, and average speeds, ideally published by Dhaka Metropolitan Police or transport authorities. Second, legal and operational processes for case generation, contestation, and appeals, because automated citations raise procedural and evidentiary questions in many jurisdictions. Third, technical transparency: model accuracy figures for plate OCR and false positive rates, and whether independent audits or third-party validations are commissioned. Local reporting suggests scalability plans, but coverage and sustained compliance remain open questions.

Dhaka's live AI-run enforcement system is an early, observable municipal use case that combines camera feeds, CV models, and automated case-generation to enforce a narrow set of traffic rules. Reported early outcomes include hundreds of auto-generated cases and anecdotal increases in compliance, while technical details and long-term impact metrics remain to be published by authorities.