COMSATS University Islamabad • BS Software Engineering
Crowd360 is an intelligent surveillance platform developed as my Final Year Project to address the challenges of monitoring large public gatherings in real time. Traditional surveillance systems rely heavily on manual observation, making it difficult for security teams to identify overcrowding, suspicious behaviour, or potential threats before incidents occur. Crowd360 combines modern web technologies with computer vision to transform live CCTV feeds into actionable insights, enabling security personnel to make faster, data-driven decisions.
The platform provides a centralized environment for event organizers and security teams to monitor live video streams, track crowd density, detect anomalies, identify weapons, and manage events through an intuitive web dashboard. By integrating AI-driven video analytics with a scalable backend architecture, Crowd360 demonstrates how intelligent surveillance can improve public safety and operational efficiency.
Managing large crowds requires continuous monitoring of multiple surveillance cameras while identifying situations that demand immediate attention. Manual monitoring is time-consuming, error-prone, and often reactive rather than proactive. The objective of Crowd360 was to build a system capable of automatically analyzing live video feeds, detecting safety risks, and presenting meaningful insights through a responsive dashboard.
Crowd360 was designed as a full-stack application with a modular architecture that separates user experience, backend services, and AI processing into independent components.
The frontend was developed using Next.js, providing a responsive dashboard for live monitoring, event management, analytics, and system administration. NestJS powers the backend, exposing RESTful APIs responsible for event management, alert processing, data synchronization, and communication between application services. AI inference is handled through a dedicated FastAPI microservice that processes live CCTV streams using YOLOv8 models for real-time object detection and crowd analysis. MongoDB serves as the primary database for storing event information, AI detections, crowd statistics, alerts, and historical monitoring data.
This architecture enables each service to evolve independently while maintaining reliable communication between the frontend, backend, AI engine, and database.
The platform continuously analyzes live CCTV footage to estimate crowd density and monitor occupancy levels across different locations. Detection results are streamed directly to the dashboard, allowing security personnel to observe crowd conditions without manual counting.
Using computer vision models, Crowd360 identifies potential security risks including weapon detection and abnormal crowd behaviour. Whenever suspicious activity is detected, the system immediately generates alerts to help security teams respond before situations escalate.
Administrators can define monitoring zones and restricted areas directly within the surveillance environment. When people enter protected regions or occupancy thresholds exceed predefined limits, the system automatically triggers notifications for immediate attention.
The application supports event-based monitoring, allowing organizers to create and manage multiple events independently. Each event maintains its own monitoring sessions, analytics, alerts, and surveillance data, making the platform suitable for conferences, concerts, sports venues, and public gatherings.
Crowd360 presents AI-generated insights through an interactive dashboard that includes live crowd statistics, density trends, alert history, and monitoring summaries. These visualizations help operators understand changing crowd conditions while providing historical data for post-event analysis.
Throughout development, emphasis was placed on building a scalable and maintainable software architecture rather than simply implementing features. The backend follows a modular service-oriented structure with clearly separated business logic, API controllers, and data access layers. RESTful APIs enable seamless communication between the frontend and AI services while ensuring consistent data flow throughout the application.
On the AI side, FastAPI acts as a lightweight inference service responsible for receiving video frames, executing YOLOv8 models, and returning structured detection results to the backend. MongoDB efficiently stores monitoring data, alerts, and event records, enabling rapid retrieval for dashboard visualizations and reporting.
The application was designed with clean coding principles, reusable components, server-side validation, structured error handling, and organized project architecture to simplify future enhancements and model integration.
Crowd360 demonstrates the practical application of artificial intelligence, computer vision, and modern full-stack development in solving real-world public safety challenges. The project showcases my ability to design scalable software systems, integrate AI services with web applications, build RESTful backend architectures, and develop responsive user interfaces that transform complex technical workflows into intuitive user experiences.
Beyond fulfilling academic requirements, Crowd360 represents an end-to-end engineering project that combines software architecture, backend engineering, frontend development, database design, and AI integration into a unified intelligent surveillance platform.