Automation of surveillance systems using deep learning and facial recognition

While it's true that humans are skilled at recognizing faces, what about computers? Machinery is transforming conventional human labor in today's technologically sophisticated world. These cutting-edge machines may be programmed to recognize faces in inconceivable ways, a process known as facial recognition or computer vision. In this study, we present a real-time system for detecting and identifying individuals in live or recorded surveillance feeds using deep learning and face recognition algorithms such as Convolution Neural Networks (CNN). The proposed real-time database integrated system is based on the VGGFace deep learning neural architecture. Transfer learning is used to retrain the model on a smaller original tailored dataset of 7500 images of 26 distinct individuals. Details regarding the creation of original tailored datasets and brief tests with various machine learning and deep learning approaches to analyze and improve the recognition accuracy of the proposed system are also reported. The proposed approach shows the highest degree of recognition accuracy by accurately recognizing each of the 26 individuals with a confidence level ranging from 78.54 percent to 100 percent, a mean average of 96 percent on real-time inputs.