Reducing road accidents requires a multi-faceted approach involving infrastructure improvement, education, and strict law enforcement. IIIT-H’s Applied AI Research Centre (INAI) is focusing on the latter with an intelligent traffic enforcement system helping generate challans as early as the pilot stage.
India has the dubious distinction of ranking number one in the total number of road accident deaths, accounting for nearly 11% of all road crash fatalities worldwide. Unfortunately it is a statistic that continues to rise despite better vehicles, stricter laws and growing public awareness. Govind Krishnan, who leads translational AI initiatives for mobility at INAI, says that the question is not why accidents happen. “We already know the answers. The real question is: How can we help solve this problem? Or more importantly, how can artificial intelligence help solve this?”
Most research projects end with a paper or a successful proof of concept. INAI’s mandate is to take promising academic research and convert it into deployable technologies that solve real-world problems. One such research effort came from Prof. Ravi Kiran S, Centre for Visual Information Technology (CVIT), whose work on AI-based helmet detection laid the foundation for VIOLA – AI-powered traffic violation detection system. The applied AI team expanded that research, added new capabilities and built it into a system production-ready system.
Prior Research Revealed Trends
INAI’s flagship road safety initiative undertaken with industry and government partners – iRASTE (Intelligent Solutions for Road Safety through Technology and Engineering) – had flagged two major problems contributing to road accidents in India. One was poor road infrastructure – potholes, damaged roads, missing lane markings and poorly maintained streets. The other was utter disregard to traffic rules and risky behaviour such as wrong-side driving, riding without helmets, triple riding and dangerous overtaking. “Road Safety Education is important,” says Govind. “But enforcement works like a charm in India. If a helmet-less rider spots a cop, he’ll immediately stop, pull out the helmet, strap it on and maybe after crossing put it away. But a cop can’t be physically present at every junction at all times of the day”. That insight became the basis for VIOLA.
Beyond CCTVs
Traffic enforcement today depends heavily on CCTV cameras. But they’re expensive to install, expensive to maintain and limited to fixed locations. Drivers soon learn where the cameras are and where they aren’t. Additionally, the ANPR technology currently being used is unable to catch all violations automatically without humans in the loop. INAI’s Engineering team wondered whether there was a better way. Their solution sounds deceptively simple. They Collaborated with Traffic Police to install inexpensive dashcams, costing around ₹8,000 to ₹10,000 each, inside existing police patrol vehicles – 4 of them to be precise. “We let the cameras record their daily rounds, fed the footage into an AI system capable of detecting traffic violations, recognising licence plates and automatically forwarded verified cases to Telangana Police’s e-Challan system,” explains Prithvi Jonnada, Operations Manager at INAI.
Unlike CCTV cameras that monitor only a single junction, dashcams travel wherever police vehicles go. This dramatically expands enforcement coverage without requiring expensive infrastructure. In this way, every patrol vehicle (equipped with a dashcam) becomes a mobile traffic surveillance unit.
What It Flags
The AI Model analyses footage to detect helmet violations, triple riding, wrong-side driving and other offences. It also reads vehicle number plates using Automatic Number Plate Recognition (ANPR), packages the evidence and submits it to the police’s existing challan generation workflow. Officers in the e-Challan generation team verify each violation before issuing a challan. “This is a mandate from the police department to have a human in the loop thus ensuring the AI acts as an assistant, not a replacement, for human judgment,” elaborates Prithvi.
Virtual Cops
With just four dashcams installed in the patrol vehicles, the system has helped generate several traffic challans within a few months. According to Prithvi, the police report that nearly 96% of the violations submitted by the AI are accurate enough to be approved. Even more encouraging, the system has become one of the largest contributors to challan generation for the Telangana Police. “Imagine what happens when this scales to hundreds of vehicles?,” exclaims Prithvi. What is distinctive is that VIOLA isn’t trying to reinvent traffic policing. Instead, it integrates seamlessly with the police’s existing e-Challan infrastructure. The dashcam becomes another source of verified violations alongside handheld devices, CCTV cameras and manual enforcement. Interestingly enough, Prithvi mentions that the police department has observed greater compliance from the general public when there is automatic flagging of traffic infractions and issuance of fines rather than via manual enforcement. “When a cop standing on the road uses a handheld device to click images of traffic violations and generates a challan, the public opposes it. But when a CCTV or a dashcam automatically recognises a violation and generates a challan, people are less frustrated or belligerent as there is no clash of the human ego.”
A Different Challenge
“We are gradually expanding the capabilities of the algorithm to include speed detection and other violations into VIOLA dashcam,” elaborates Govind. The implications extend beyond issuing fines. The same dashcam footage that captures traffic violations can also capture road conditions, identifying potholes, damaged roads and missing lane markings, thus feeding valuable information back to civic authorities.
For the team, VIOLA represents more than a successful AI application. It illustrates how IIIT-H’s innovation pipeline moves ideas from research to real-world deployment. The underlying AI research that originated at CVIT was developed into a deployable solution through INAI’s applied AI Research Center, and received support through iHub-Data, IIIT-H’s technology innovation hub focused on converting research into societal impact.
The team is now in discussions with police departments beyond Telangana and exploring partnerships that could support wider deployment. As the solution matures, pathways for broader adoption and commercialization are also being explored through the institute’s innovation ecosystem. “People don’t expect an academic institution to build a complete industrial-grade product that is deployed and generating real impact. We want to change that,” says Dr. Ganesh Yalla, CEO of iHub-Data.
Sarita Chebbi is a compulsive early riser. Devourer of all news. Kettlebell enthusiast. Nit-picker of the written word especially when it’s not her own.