International Journal of Advanced Research in Computer and Communication Engineering

Abstract: The implementation and design of a hybrid detection framework driven by AI that can analyse objects and faces in real time using a single web-based system is introduced in this paper. The proposed architecture uses a dual-model pipeline that combines the speedy object detection capabilities of You Only Look Once, version 7 with the facial recognition accuracy of DeepFace. A full-stack Flask application serves as the foundation for secure user interaction, camera management, and live data visualization. The framework facilitates multi-camera connectivity, dynamic object selection, and instant user enrollment through interactive face capture. Effective object monitoring and human identification in a range of environmental conditions are ensured by the detection and recognition modules operating simultaneously on live video streams. All user credentials and logs are securely maintained using encrypted authentication techniques and Structured Query Language Lite. Additionally, by allowing real-time updates of facial datasets without server outages, the system improves scalability and flexibility. The experimental evaluation demonstrates that the hybrid model consistently provides high recognition accuracy while maintaining low processing delays, making it suitable for real-time applications such as automated attendance, intelligent monitoring, and security-driven surveillance. Its combined structure allows the system to handle both object detection and identity recognition within a single workflow, avoiding the limitations of using separate models. By merging these capabilities, the framework delivers a balanced solution that improves reliability, strengthens security, and ensures smooth real-time operation. Overall, the developed system creates an integrated platform that aligns efficiency, adaptability, and practical usability for diverse AI-based environments.

Keywords: Hybrid Detection Framework, YOLOv7, DeepFace, Real-Time Recognition, AI-Driven Face Analysis

Downloads: | DOI: 10.17148/IJARCCE.2025.141188