OmniSight - AI Pioneered Intelligent Crowd and Safety System

Problem Statement

Venue management is an increasingly complex challenge due to:

• Inefficient crowd control leading to congestion
• Delays in emergency response
• Security risks from unidentified objects
• Long wait times and slow service causing customer dissatisfaction.

---

Introduction

Managing large venues efficiently is challenging due to congestion, security risks, and long wait times.
Our AI-powered venue management system integrates video analytics with the Nx Developer Toolkit to:

• Optimize crowd control
• Detect fire and security threats
• Automate food stall queue management & resource allocation
• Wait time analysis & dynamic queue management for events
• Implement dynamic pricing & marketing

This solution ensures real-time insights, enhanced security, and seamless customer experience.

---

Features

Our AI-powered venue management system enhances efficiency, security, and customer experience with real-time analytics and automation.

Crowd Monitoring & Control

• Real-time people detection using YOLOv8 (People Detection)
• Heatmap-based congestion analysis using a grid-based spatial mapping algorithm
• Predictive modeling to prevent overcrowding

Fire & Smoke Detection

• AI-driven fire and smoke detection using YOLOv8 (Custom Trained)
• Instant alerts to security teams and venue management
• Integration with emergency response systems for faster action

Threat Detection

• Identification of suspicious objects such as unattended bags, knives, and hazardous items using YOLOv11s
• Real-time alerts and notifications for security personnel
• Automated tracking of flagged objects using AI-powered object tracking

Wait Time Analysis

• Uses Exponential Moving Average (EMA) for queue size prediction
• Real-time wait time estimation based on queue length and predictive AI modeling
• Provides estimated service times to improve customer experience

Dynamic Queue Management

• AI-based real-time queue optimization for food stalls and event entry
• Smart allocation of waiting areas to reduce congestion
• Predictive adjustments to queue formations based on peak and off-peak hours

Automated Food Stall Queue & Resource Allocation

• AI-driven staff allocation algorithm to manage cashiers, cooks, and servers
• Predictive order flow analysis to prevent food stall bottlenecks
• Automated load balancing to ensure efficient food service

Dynamic Pricing

• AI-powered real-time pricing adjustments based on demand fluctuations
• Automated price increases during high demand and discounts during low demand
• Ensures revenue maximization while balancing customer satisfaction

Targeted Marketing

• AI-driven food detection using YOLOv11s (Custom Dataset) to track demand
• Personalized promotions based on real-time food stall activity
• Automated special deals and discounts for low-demand items

---

This system ensures efficient crowd flow, reduced wait times, enhanced security, and optimized business operations.

AI Models Used

Model	Purpose	Technology
Fire & Smoke Detection	Detects fire hazards	YOLOv8 (Custom Trained)
Crowd Management	Monitors crowd density and flow	YOLOv8 (People Detection)
Food Detection	Recognizes food items at stalls	YOLOv11s (Custom Dataset)
Dangerous Object Detection	Identifies objects like suitcases and knives	YOLOv11s

---

🛠️ Tech Stack

Backend: Python (Flask) Frontend: React, Tailwind CSS
AI Models: YOLOv8, YOLOv11s (Custom trained)
Nx Integration: Nx Developer Toolkit

---

Algorithmic Approach

Real-Time Heatmap Generation

A grid-based spatial mapping algorithm tracks people’s movement, creating live heatmaps.

Algorithm:

1. Capture real-time video feed and detect people.
2. Map detected person’s centroid (cx, cy) onto a grid.
3. Update heatmap based on position frequency.

row = cy / cell_height
col = cx / cell_width

Exponential Moving Average for Queue Smoothing

To prevent erratic queue size fluctuations, an Exponential Moving Average (EMA) algorithm smooths the data:

Algorithm:

1. Measure real-time queue size.
2. Apply EMA smoothing to reduce sudden fluctuations.
3. Use a smoothing factor α (0 < α ≤ 1) to adjust responsiveness.

EMA_Size = α * Queue_size + (1 - α) * EMASize

Wait time estimation algorithm

Wait time is estimated dynamically.

Algorithm:

1. Measure queue length in real-time.
2. Multiply queue length by the base service time per customer.
3. Provide an estimated waiting time for customers.

Wait_Time = Queue_length * Base Time

Dynamic Staff Allocation Algorithm

The system optimizes staff allocation based on real-time queue size.

Algorithm:

1. Compute required staff using EMA queue size.
2. Assign cashiers, cooks, and servers dynamically.
3. Ensure minimum and maximum staff limits.

required_staff = max(1, EMA_Size / 3)
cashiers = min(5, required_staff / 3 + 1)
cooks = min(6, required_staff / 2)
servers = required_staff - cashiers - cooks

Dynamic Pricing

The system dynamically adjusts pricing and marketing based on demand.

Algorithm:

1. Monitor food demand.
2. Adjust pricing based on thresholds:
   • Increase price during high demand.
   • Decrease price when demand is low.

if demand > HIGH_THRESHOLD:
    price *= 1.2  # Increase price for high-demand items
elif demand < LOW_THRESHOLD:
    price *= 0.8  # Offer discounts for low-demand items

Targeted Marketing

Marketing messages are generated based on real-time food stall demand.

Algorithm:

1. Check food demand levels.
2. Display different marketing messages based on sales.

if count >= HIGH_DEMAND_THRESHOLD:
    message = f"{food.capitalize()} is selling fast! Get yours before it's gone!"
elif count == MEDIUM_DEMAND_THRESHOLD:
    message = f"{food.capitalize()} is a customer favorite! Order now!"
elif count <= LOW_DEMAND_THRESHOLD:
    message = f"Special deal on {food.capitalize()}! Limited time offer."