Guardio: A Home Intrusion Detection System

Abstract

The increasing number of intrusions in homes or private property is a huge issue which prevails in the society. The intrusions lead to loss of personal items and sometimes causes threats to human life specifically in a residential area. The project focuses on developing an intrusion detection system that could detect intruders using their behavior and generating a description of the intruder. A hybrid model was designed which combines You Only Look Once (YOLO) object detection framework and Natural Language Processing (NLP) for the intruder detection and description generation. The application processes data which consists of intrusions under several lighting conditions and preprocessing approaches such as feature selection and outlier removal would be done for the use of fine tuning and accuracy assurance. The accuracy was significantly good based on the executed test results. The accuracy was reasonable in determining the person objects indicating no false negatives. However, objects were identified with incorrect captions. The accuracy would be improved and obtained as a numeric when the model is developed for the MVP.