CrowdWise | Devpost

Problem Statement: What challenge does my project solve?

Every year, stampedes at Indian large gatherings , festivals, and railway stations kill hundreds of people. In January 2025, 30 people died and 60 were injured at Kumbh Mela, Prayagraj. In 2008, 147 died at Chamunda Devi temple in Jodhpur due to stampede situation.

Current crowd management relies entirely on manual headcounts and radio communication. By the time a dangerous situation is spotted, it is already a crisis. There is no early warning, no prediction, and no automated response. Current approaches are reactive, manual, and inadequate. The failure is not human error. It is the complete absence of real-time crowd intelligence.

Solution Overview

CrowdWise is a real-time AI crowd management system that shifts public safety from reactive crisis response to proactive prevention.

When a crowd turns dangerous, human reaction is too slow. CrowdWise detects, predicts, reroutes, and instructs all in under 3 seconds before panic sets in.

The system works in four stages:

Detect — YOLOv8 neural network counts people from live camera feeds every second across all venue zones
Predict — AI forecasts crowd density 15 minutes ahead and fires alerts before the situation becomes critical
Reroute — Dijkstra graph algorithm calculates the safest alternate crowd path instantly, producing three ranked options for the operator
Instruct — Digital signboards update automatically in Hindi and English with safety instructions and estimated wait time within 3 seconds

Event organizers upload their venue blueprint before the event. The system analyzes the layout, divides it into zones, calculates safe capacity for each, and maps all possible routes automatically. No manual configuration needed.

Key Features:

Real-time person detection using YOLOv8 with live bounding boxes and confidence scores
Multi-zone crowd monitoring - each zone shows count, capacity percentage, status, and wait time updated every second
15-minute predictive warning system that alerts operators before density reaches critical levels
SmartRoute engine producing three distinct ranked routes based on real-time zone densities operator can select manually or let AI decide
Multilingual digital signboard in Hindi and English updating automatically with crowd state

State	Hindi	English
SAFE	स्वागत है — कृपया शांति से आगे बढ़ें	Welcome — Please proceed calmly
MODERATE	भीड़ बढ़ रही है — कृपया धीरे चलें	Crowd building — Please move slowly
DANGER	खतरा — दिखाए गए मार्ग का पालन करें	Follow the recommended route on display
EMERGENCY	आपातकाल — निकटतम निकास की ओर जाएँ	Move to nearest exit immediately

Live crowd density heatmap color-coded from green to red to magenta as density increases
Crowd flow map showing active route with directional arrows
Blueprint upload system auto-generating zone layouts for temples, stations, malls, and stadiums
Prototype includes a simulation mode cycling Safe to Emergency automatically designed for demonstration purposes

Technologies Used

Python and Flask — backend and REST API
YOLOv8 (Ultralytics) and OpenCV — real-time person detection
NumPy — polynomial regression crowd prediction
Custom Dijkstra algorithm — graph-based route optimization
Chart.js and HTML5 Canvas — live graphs, heatmap, flow map
JavaScript — single-page dashboard

Full deployment roadmap includes TensorFlow/PyTorch LSTM, React Native mobile app, AWS cloud infrastructure, and IoT people counter integration.

Target Users

Control Room Operators Role: Monitor crowds, make decisions, activate protocols Interface: Web-based control dashboard Key Needs: Single-screen view, color-coded alerts, quick-access protocols

Field Officers Role: Execute ground responses, manage crowd movement Interface: Mobile application (Android/iOS) Key Needs: Push notifications, GPS assignment, hands-free operation

General Public Role: Follow guidance, move safely Interface: Digital signboards and audio announcements Key Needs: Large signage, simple instructions, multilingual support

Secondary Stakeholders

Event Organizers — Rent system for festivals and concerts

Religious Trusts — Deploy at temples and processions to protect devotees

Mall Operators — Subscription for permanent crowd flow management

Stadium Owners — Event-based activation for entry and exit control

Emergency Services — API integration for early alert positioning

Smart City Initiative — Flagship AI governance and public safety project

Social Impact Statement

India loses hundreds of lives every year to preventable stampedes. The technology to stop this already exists computer vision, graph algorithms, predictive modelling. CrowdWise combines them into a system specifically designed for India's highest-risk public gatherings. If successfully implemented, CrowdWise has the potential to prevent stampede-related deaths, reduce emergency response costs, and restore public confidence in attending large gatherings. Beyond Mumbai, the system can serve as a model for crowd safety in dense urban environments worldwide.