A scene that would have been hard to imagine just months ago is now becoming increasingly common on Dhaka's streets; motorists patiently waiting at a red light on an almost deserted road.
At 12:55am on Sunday, Mehedi Hasan Shanto was travelling by motorcycle from Purana Paltan to Mirpur when the rider suddenly stopped at the traffic signal near Hotel InterContinental, despite there being no other vehicles at the intersection.
Surprised by the sudden brake, Shanto asked why they had stopped. The rider simply pointed towards an AI-powered traffic camera overlooking the junction. "If we cross now, we'll be fined," the rider replied.
For Shanto, the moment reflected a remarkable shift in road behaviour.
"I never imagined AI cameras could bring such a rapid change in traffic discipline in Bangladesh," he told BSS.
The transformation is becoming increasingly visible across the capital. Since the introduction of AI-powered traffic surveillance at major intersections, many motorists have begun respecting traffic signals, stopping behind designated stop lines and keeping zebra crossings clear for pedestrians.
DMP Additional Commissioner (Traffic) Md. Anisur Rahman described the AI initiative as a milestone in Dhaka's traffic management.
"The biggest achievement of AI cameras is the visible improvement in traffic discipline. Drivers now obey traffic signals more consistently and stop before zebra crossings,” he said.
“We currently operate 117 cameras, including 37 PTZ cameras, and the network will be expanded gradually to bring the entire capital under AI-based traffic monitoring," he added.
The transformation is equally visible during the city's busiest hours.
At Karwan Bazar, one of Dhaka's busiest intersections, buses, motorcycles and private cars now routinely stop several feet before the stop line when the signal turns red, allowing pedestrians to cross safely through zebra crossings without obstruction.
Only a few months ago, drivers frequently ignored traffic signals, occupied pedestrian crossings and crossed stop lines with little regard for traffic rules.
Pedestrian Ruman Mahmud said crossing the road during office hours used to be extremely difficult.
"Earlier, vehicles would occupy the zebra crossing, and motorcycles even used the footpath. Now, drivers are much more careful because they know the AI cameras are watching," he said.
Police officials say the country's AI-based traffic management initiative has been implemented at remarkably low cost.
The project initially began with an investment of only Tk 60 lakh, while the total expenditure so far has remained below Tk 1crore.
Currently, 117 AI-enabled cameras are operating at 19 major intersections, including Abdul Gani Road, Matsya Bhaban, Kakrail Mosque, Sugandha, State Guest House Jamuna, Police Bhaban, Hotel InterContinental, Bangla Motor, Sonargaon Crossing, Bijoy Sarani, Jahangir Gate, Mohakhali Bus Terminal, Shahbagh, Science Lab and Farmgate.
Authorities plan to gradually expand the network across the capital, although budgetary and infrastructure constraints prevent citywide implementation at once.
The AI system uses Pan-Tilt-Zoom (PTZ) cameras capable of detecting vehicle number plates from long distances, tracking moving vehicles and capturing high-resolution footage.
Beyond red-light violations, the cameras can identify wrong-way driving, stop-line violations, illegal occupation of zebra crossings, riding motorcycles without helmets, driving without seat belts, using mobile phones while driving, illegal parking and unauthorized use of VIP lights.
However, police stress that no case is filed solely on the basis of AI detection. Traffic officials first manually verify each recorded incident.
After confirmation, a notice is sent to the registered owner by letter and mobile message. Around 15 to 20 seconds of video footage are preserved to allow review if anyone disputes the allegation.
According to police, the objective is not punishment but ensuring safer roads through transparent and technology-driven enforcement.
More than 1,500 traffic cases have already been processed using AI-assisted monitoring.
Officials acknowledge that the technology still faces limitations. Heavy rain, night-time conditions and traffic congestion sometimes make number plate recognition difficult.
Many vehicles also carry faded, damaged or unreadable number plates, complicating identification. Police have now begun taking legal action against such vehicles.
Recently, a motorcyclist who intentionally obscured part of his registration plate was identified through the system and sentenced to imprisonment.
Emergency vehicles such as ambulances and fire service units remain exempt from AI-based enforcement.
The technology has also changed the daily responsibilities of traffic police.
Instead of running into busy roads to stop vehicles or conducting frequent roadside document checks, officers can now focus on maintaining overall traffic flow while reducing risks to themselves.
Traffic Sergeant A Shukur, who serves at the Hotel InterContinental intersection, said the new system has significantly improved officer safety.
"We no longer have to rush into traffic or physically stop vehicles. The risk of being hit by a car has reduced considerably. If unauthorized rickshaws can also be kept off the main roads, traffic discipline will improve even further," he said.
