With AI assistance, Bengaluru techie turns helmet into traffic watchdog

Bengaluru: On the city's chaotic roads, where red lights are often treated as suggestions and wrong-side driving is routine, a 27-year-old software engineer has decided to fight back using code instead of confrontation. Pankaj Tanwar, a city-based techie, has fitted his helmet with an Artificial Intelligence-enabled camera that automatically detects traffic violations and reports them directly to traffic police, complete with evidence.Tanwar's experiment came to public attention after he posted a video and description of the system on X. "I was tired of stupid people on the road, so I hacked my helmet into a traffic police device," he wrote, adding that the system flags violations in near real time and sends proof, along with location and vehicle number, to police.
Originally from Rajasthan, Tanwar moved to Bengaluru for work and is currently employed as a software engineer. The Kadubeesanahalli resident said the idea was born out of daily frustration. "I kept seeing violations around me. Arguing with drivers was pointless, so I looked for a technological solution," he said.Tanwar discovered that Bengaluru traffic police already accept citizen-reported violations through their app, email and WhatsApp, provided the complaint includes supporting evidence such as photographs, videos, and location details. That became the foundation for his helmet-mounted system.The setup uses a Logitech webcam attached to the helmet and connected to a Raspberry Pi, a compact computer that processes live video as he rides. Each frame is analysed by AI models to detect four common violations: riding without a helmet, jumping dividers, running red lights, and driving on the wrong side of the road.To reduce errors, Tanwar has added a second layer of verification. Detected violations are cross-checked using multiple AI models, including OpenAI's GPT and Google's Gemini, before being flagged as valid. "The system checks frames over a five-second window. If the confidence score crosses 96%, the report is sent automatically," he said.Once verified, an email is sent to police with still images, short video clips, the vehicle's number plate, and GPS coordinates. Police then issue a challan, without any further input from Tanwar.Built largely from spare parts, the device cost him under Rs 4,000. With scale, he estimates the cost could drop to around Rs 2,000. Powered by a standard power bank, it runs for six to seven hours and can be adapted for cars, taxis or trucks. "The idea is scalable monitoring without installing expensive cameras everywhere," he said. Tanwar said traffic police have reached out to him and scheduled a meeting. Joint commissioner of police (traffic) Karthik Reddy told TOI he would review the post and speak to the techie to better understand the system.