Near-miss video studies that show cities which crossings endanger pedestrians
TL;DR Conflictcam runs software over video from junction cameras a city already owns and counts the near misses between pedestrians and vehicles. You get a ranked list of crossings to fix and a before-and-after test once they're fixed. Municipal bodies, smart-city companies and road-safety consultants pay per study.
Walking is one of the most dangerous ways to get around in India. MoRTH counted 1,68,491 road deaths in 2022[1], and pedestrians made up 19.5% of them[2], which works out to roughly 32,800 people in a single year by our arithmetic.
Cities usually decide where to act from crash records. Those records arrive late, depend on what police wrote down at the scene and only flag a junction after people have died there. A crossing where drivers miss pedestrians by half a second every evening looks perfectly safe on paper until the first fatality. Engineers have a better signal in the near miss, but measuring it has meant manual video coding that few city budgets will fund.
The seed pitched augmented-reality warnings on pedestrians' devices, and I don't think that works here. Almost nobody in India walks around wearing AR glasses, and a phone alert pulls a pedestrian's eyes down at exactly the wrong moment. The people who can change a crossing are the ones who own the road, so Conflictcam serves them.
The city shares a week of footage from cameras at a junction it's worried about. Conflictcam tracks every pedestrian and vehicle, measures how close each pair came using standard conflict metrics such as post-encroachment time and flags the worst patterns - turning traffic cutting through crowds, people stranded mid-road and drivers who never yield at a zebra. Indian researchers have already shown these measures work on mixed traffic at unsignalised crosswalks[3].
The report ranks crossings, shows short anonymised clips of typical conflicts and suggests standard fixes for engineers to judge. After a refuge island or a signal-timing change a second study checks whether conflicts actually fell. Faces and number plates are blurred before anyone sees a clip.
Pedestrians benefit, but the payer is whoever owns the junction, be it a municipal corporation, a smart-city company, a development authority or the traffic police. Road-safety audit consultants are a second buyer, adding the study to work they already bid for. I found no public price for near-miss studies in India. My test assumption is ₹1.5–3 L per junction for a seven-day study, with a cheaper rate for corridors of ten or more. Government buying often runs through GeM[4], which takes registration and a lot of patience.
The cameras are already up. Indian traffic platforms sell live junction feeds and automated violation detection to cities like Bengaluru[5], so footage exists without new hardware. MoRTH is pushing electronic accident reporting through e-DAR[1]. The DPDP Rules notified in November 2025[6] make the duties on processing video of the public clearer than they were.
There's no public figure for the Indian market in road-safety video analytics or near-miss studies. The firmest number is the size of the harm, and about one in five road deaths is a pedestrian[2]. The closest foreign comparable, a US intersection-safety firm, says its intersection system is live in the US, Canada and the GCC[7], which shows that city agencies elsewhere pay for this kind of analysis. Treat that as a precedent, not a market size.
Indian traffic-video vendors focus on enforcement, spotting helmet-less riders, red-light jumps and speeding and then issuing e-challans[5][8]. That US intersection-safety firm does near-miss insights and warnings at intersections for agencies abroad[7]. Academic teams in India run conflict studies, but as research and not as a service. Nobody we found in India sells pedestrian near-miss studies to cities as a repeatable product.
| Difficulty | Moderate (46/100) |
| Starting capital | ₹55 L–1.1 Cr |
| Time to launch | 6–9 months |
Tracking people reliably in dense, low-angle, night-time Indian footage is the technical hurdle, and models trained on Western intersections miss a lot. The harder part is getting footage at all. Camera networks sit with police or city vendors, and each one needs a data-sharing agreement before a single frame moves.
No sector licence covers near-miss video analysis, but you can't touch a frame without permission.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
AI counts conflicts, but it shouldn't decide what to build. Check its counts against a manual sample at every new junction, and leave the design to an engineer.
Want to work out the AI layer for Conflictcam? Talk to Yash →
Here's what can go south -
I think this suits a founder with a traffic-engineering or computer-vision background who already knows how city agencies buy. If you're hoping for consumer scale then it's the wrong idea. Prove two things first - that one city will hand over a week of footage within 60 days of asking, and that your conflict counts match a manual count closely enough for an engineer to sign off on them.
Plan for roughly ₹55 L–1.1 Cr, covering model work for Indian traffic, pilot studies and a long government sales cycle.
No sector licence, but you need a written footage-sharing order from whoever controls the cameras and must follow the DPDP Act and Rules 2025.
Yes, cities everywhere want to know which crossings are risky. Check local data protection rules on video and how your city buys road-safety studies.
The free Conflictcam validation kit has the line-by-line capital breakdown, the effort scoring, the data-sharing and privacy checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for city officials and consultants, and 5 assumptions to test. Download the kit →
Thinking about building Conflictcam or something close to it? The make-or-break work is getting a police or city camera owner to sign a data-sharing agreement, and I'm happy to help you plan that first ask. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.