CCTV AI-Based System with Custom Object Detection and Video Upscaling Model

This chapter presents an affordable and secure CCTV system that integrates custom object detection and video upscaling models. The system aims to enhance the security capabilities of CCTV systems while remaining cost-effective and accessible. It incorporates an alert mechanism to promptly notify security personnel via popups, sound, and email notifications. The custom object detection model, based on Yolov5 architecture, accurately detects potential threats by leveraging convolutional neural networks and pooling layers. The video upscaling model, utilizing Real-ESRGAN architecture, improves the resolution and quality of surveillance videos. Custom datasets were curated for both models, including high-resolution videos with motion and firearm instances. The system's user interfaces were developed using Kivy and Tkinter, offering intuitive interactions. With features such as object detection, alerts, and an “Enhance” option, the system provides an effective security solution. The technical analysis of the models, dataset evaluation, and modifications made to ensure secure communication are discussed. This affordable and secure CCTV system with custom object detection and video upscaling models offers a valuable contribution to video surveillance, benefiting users with limited budgets while maintaining a high level of security.