IIIT-H’s Vahan Eye cracks India’s hand painted truck plate puzzle
HYDERABAD: When commercial automatic number plate recognition (ANPR) systems failed to decode India’s hand-painted truck plates, researchers at IIIT Hyderabad turned the challenge into an innovation opportunity. Their in-house team at iHub Data developed Vahan Eye, a low-cost, field-deployed system now monitoring sand transport in Telangana.
‘Truck art’ the vibrant “Horn Ok Please” and “Use Dipper at Night” signs is integral to Indian highways, often extending to hand-painted registration plates. But this creativity poses a hurdle to standard ANPR systems, which rely on uniform fonts and layouts.
Tackling a unique number plate challenge
When the Telangana IT Department approached IIIT-H for a solution for the Telangana State Mineral Development Corporation (TGMDC), the task was specific: design an affordable, robust system to track sand trucks and curb illegal mining.
“Typical plates are easy to detect,” said Dr Veera Ganesh Yalla, CEO of iHub-Data and adjunct faculty at IIIT H. “But truck plates here are hand-painted, inconsistent and highly variable.” The diversity of styles rendered commercial systems both ineffective and costly.
Building smart, not from scratch
Most commercial ANPR systems cost several lakh rupees per camera. The iHub-Data team built on research led by Prof Ravikiran Sarvadevabhatla at the Centre for Visual Information Technology, where a prototype existed. They rebuilt the handwritten character recognition component and integrated it into an open-source platform. “Anyone can plug in our module without rewriting their system,” said Dr Yalla.
Real-world deployment in Telangana
The Vahan Eye system was piloted at Chityal on the Vijayawada–Hyderabad highway. Cameras and sensors track trucks entering the state and verify them against a whitelist of 40,000 authorised vehicles. Since September, it has operated continuously, adapting to challenges such as low light and plate obstructions during festivals.
From prototype to policy tool
Built by a five-member team using deep learning models such as YOLO and RF-Detr, the system is now being adapted for the Police Department to detect two wheeler traffic violations. “Our IP lies in solving handwritten license plate recognition,” said Dr Yalla, adding that the goal is to make scalable, affordable ANPR technology accessible to public agencies.
