Crowd Management During Sawan: How AI-Powered CCTV Keeps Temple Towns Safe
Published on 05 Aug 2026

Introduction

It’s 4 AM in a temple town somewhere in North India. The gates haven’t opened yet, but the queue outside already stretches past the last streetlight. By noon, this town’s population will have quietly tripled. By evening, the parking lots will be full, the lanes leading to the sanctum will be shoulder to shoulder, and somewhere in the crowd, an elderly kanwariya will be looking for a bench that doesn’t exist.

This is Sawan. And every year, the story repeats itself, not because temple administrations don’t care, but because they’re trying to manage a modern crowd problem with a pre-modern toolkit: a few dozen guards, some barricades, and hope.

We think the toolkit needs an upgrade. Here’s what that actually looks like, hour by hour.

5 AM — The Gates Open, and So Do the Blind Spots

The first rush is always the hardest to predict. Nobody knows exactly how many people are inside the complex at any given moment, not the administration, not the police, not the volunteers coordinating water and medical camps.

This is where footfall counting changes the entire equation. Cameras that are already installed for security, the same CCTV infrastructure sitting on poles and gates can be turned into real-time people-counters. Instead of guessing whether the complex is at 60% or 160% capacity, administrators get a live number. Not an estimate from last year’s Sawan. A number, right now, updating as people walk in and out.

That single number is the difference between “let’s see how it goes” and “we need to slow entry at Gate 3 before it becomes a problem.”

9 AM — The Moment Before It Becomes a Headline

Every stampede story in the news starts the same way: a bottleneck nobody saw building until it was too late. A narrow lane, a sudden surge, a few thousand extra people who arrived in the last twenty minutes.

Stampede prevention– isn’t really about reacting faster once a crowd crush starts, by then it’s already a crisis. It’s about catching the density curve early. When camera-based analytics flag that a specific lane or gate is approaching unsafe crowd density, not “crowded,” but a measurable, rising number teams can act while there’s still room to act: reroute a queue, open a second lane, pause entry for ten minutes.

It’s the same technology as the footfall counter, just pointed at a different question. Not “how many people are here” but “how many people are here, in this ten-meter stretch, right now.” That’s the number that actually prevents tragedies.

11 AM — Who’s Actually on Duty?

Behind every well-managed Sawan crowd is an even bigger, less visible crowd: volunteers, medical staff, security personnel, and municipal workers, all rotating through long shifts across a sprawling temple complex.

Manual attendance registers don’t scale to this. A volunteer who signed in at 6 AM at the east gate might genuinely be needed at the west gate by 11, but if nobody can confirm who’s actually present and where, planning tomorrow’s shift is guesswork.

Facial recognition attendance solves this quietly, in the background. People are marked present the moment they pass a camera, no queues, no registers, no double-checking. For an event coordinator managing hundreds of temporary staff across a multi-day mela, that’s not a minor convenience. It’s the difference between knowing your actual on-ground strength at any hour and finding out you were short-staffed only after something went wrong.

1 PM — The Traffic Jam Nobody Warned You About

Ask any local shopkeeper what actually breaks down first during Sawan, and most won’t say the temple gates. They’ll say the parking.

Buses, private cars, two-wheelers, all converging on a town that has maybe a tenth of the parking it needs, on roads that weren’t built for this. When lots fill up unannounced, vehicles start parking wherever they can, and what should be a footpath becomes a bottleneck too.

Parking utilisation monitoring turns this from a mystery into a managed system. Cameras track occupancy across designated lots in real time, so authorities know exactly which lots are full and which still have room and can direct incoming traffic accordingly, before it piles up at the entrance. Pilgrims spend less time circling for a spot. Local roads stay clearer. And the town’s traffic police get a live map instead of a walkie-talkie full of guesses.

3 PM — The Person Everyone Forgets to Design For

Here’s what usually happens when something goes wrong for a pilgrim mid-crowd: a toilet block runs out of water, a lane goes dark after sunset, someone’s bag goes missing, a loudspeaker outside a rest area won’t stop blaring at 2 AM. In a normal year, none of this gets reported. There’s no one to tell, and even if there were, nobody has time to stand in another queue just to file a complaint.

This is the gap Enalytix’s Citizen Feedback App is designed to close. It’s a simple, category-based reporting screen: Toilet/Sanitation Issues, Transport/Traffic, Overcrowding/Safety, Cleanliness/Waste, Lost Items/Theft, Drinking Water Problem, Lighting/Electricity, Loud Noise/Disturbance, and a catch-all Other Complaint.

Any pilgrim taps the category that matches their problem and it goes straight to the team that can actually act on it, no standing in line, no chasing down a volunteer, no complaint lost in the noise of the crowd. It’s built for every pilgrim in the crowd, but it matters most for the ones least equipped to chase down help on their own elderly visitors especially, who are the least likely to track down a volunteer when a toilet is unusable or a lane is unsafe after dark, and least likely to have the patience to file a complaint through five layers of bureaucracy.

A tap-and-report system means their problem reaches someone the moment it happens, not after a family member notices, not after it becomes a bigger issue.

For the administration, it’s the same principle as everything else in this system: turning something invisible into something they can see and act on.

Overcrowding and safety complaints flow into the same operational picture as the density data from the gates. Lost item reports can be cross-checked against the same cameras tracking footfall. Nothing about the crowd stays unreported just because no one had the time to walk over and say something.

By Nightfall, It’s Not Guesswork Anymore

None of this requires temple towns to rebuild their infrastructure from scratch. The cameras are usually already there, watching gates and lanes for security. What changes is what those cameras are asked to do count instead of just record, flag density instead of just capture footage, recognize a face for attendance instead of only for surveillance.

That’s the real story of crowd management during Sawan: not new hardware, but smarter use of what’s already watching. Footfall counting tells you how many. Stampede prevention tells you where it’s getting dangerous. Facial attendance tells you who’s on the ground. Parking monitoring tells you where the vehicles are. And the Senior Citizen App makes sure no one gets lost in a system built for millions.

Put together, it’s not a surveillance story. It’s a safety story, one where a temple town that welcomes lakhs of pilgrims a day can do it without leaving anyone, from the frontline volunteer to the elderly kanwariya, to chance.

Enalytix turns existing CCTV infrastructure into intelligent, insight-driven systems helping organizations manage crowds, safety, and operations without ripping out what’s already working.