AI-Based CCTV Analytics
Real-time AI-driven face and vehicle recognition for enhanced security and operational efficiency in APSCSCL Warehouses.
APSCSCL Smart Surveillance Hackathon
The APSCSCL Smart Surveillance Hackathon (April 2025 Challenge) sought innovative AI solutions for video analytics using existing CCTV infrastructure. Key objectives included enhancing warehouse security, improving operational efficiency, and enabling real-time monitoring through advanced object detection, recognition, and event analysis. Our project directly addresses these critical needs.
Our Innovative Solution
👤 Face Recognition
We employ ArcFace, a cutting-edge deep learning model for highly accurate face feature extraction, integrated via the InsightFace library. This system reliably identifies authorized personnel in real-time, enhancing access control and enabling precise tracking of individuals within warehouse premises.
🚗 Vehicle Recognition & Tracking
Our vehicle analysis module uses YOLOv8 for robust vehicle detection. Subsequently, PaddleOCR (with OpenALPR as an alternative) performs license plate recognition. This dual system allows for automated logging of vehicle entries/exits, tracking movement, and flagging unauthorized vehicles, streamlining logistics and bolstering site security.
System Architecture
Our system architecture is designed for scalability and real-time processing, integrating multiple AI models and data streams efficiently.
System Pipeline: CCTV Feed Ingestion → Frame Preprocessing → Parallel AI Modules (Face Recog. & Vehicle Recog.) → Kafka Message Bus → Data Aggregation & Storage → Real-time Dashboard & Alerting System.
Solution Demo
Witness our AI-powered surveillance system in action. This demonstration highlights key functionalities and real-world application.
Analytics Insights
Visualizing key metrics and trends identified by our AI system over time. (Live chart replaced with placeholder)
)
Meet the Team & Our Tech Stack
Our Innovators
Technologies Powering Our Solution
Connect With Us
Explore our project further, contribute, or get in touch for collaborations.
Connect with team members on LinkedIn: