The Application of Virtual Fencing in Home Security Systems

To increase the security of homes and bring relief to homeowners when they are away from it, Security Systems that may be able to perform real-time monitoring and detection of intruders breaching the bounds of their property could be solved by implementing the concept of Virtual Fences in Computer Vision using Machine Learning. This paper covers previous implementations of Virtual Fencing while emphasizing Object Detection methods that are required to perform such a task. Explores the dichotomy between single stage and two stage detectors and compares their architecture, performance, and applications. It examines various datasets that may be used to train object detection models and performance metrics that could be used to analyse and benchmark the capabilities of such models. A proof of concept is developed using the YOLOv8 model and a custom Virtual Fencing Logic along with a comparison between performance of YOLOv8 and YOLOv9 variants.