VIGILE AI: A HYBRID EDGE-AI AND VISION-LANGUAGE FRAMEWORK FOR REAL-TIME INTRUSION DETECCTION AND INTELLIGENT THREAT ASSESSMENT

This paper presented VigileAI, an intelligent surveillance framework combining Edge AI and Vision-Language Models for real-time threat assessment. The proposed system successfully integrates YOLOv8-based intrusion detection, crowd analysis, evidence generation, and Gemini-powered scene interpretation into a unified architecture. Experimental results demonstrate that the framework provides accurate detection, low latency, and enhanced contextual awareness. The study highlights the potential of combining computer vision and multimodal reasoning to create next-generation intelligent surveillance systems.

[1]              J. Redmon et al., "You Only Look Once: Unified, Real-Time Object Detection," CVPR, 2016. [2]              G. Jocher et al., "Ultralytics YOLOv8 Documentation," Ultralytics, 2024. [3]              N. Dalal and B. Triggs, "Histograms of Oriented Gradients for Human Detection," CVPR, 2005. [4]              S. Sultani et al., "Real-World Anomaly Detection in Surveillance Videos," CVPR, 2018. [5]              A. Radford et al., "Learning Transferable Visual Models From Natural Language Supervision," ICML, 2021. [6]              Google DeepMind, "Gemini: A Family of Highly Capable Multimodal Models," 2024. [7]              OpenCV Development Team, "OpenCV Documentation," 2024. [8]              Pressman, R. S., Software Engineering: A Practitioner's Approach, 2020. [9]              Russell, S. and Norvig, P., Artificial Intelligence: A Modern Approach, 2021. [10]           Goodfellow, I., Bengio, Y., and Courville, A., Deep Learning, MIT Press, 2016.