A Review on: A Lightweight DCCN System for Real-Time Surveillance that Detects Abandoned Objects

In the context of video surveillance, abandoned object detection has become all the rage in recent years. Hence, several methods have been suggested to automatically monitor both public and private spaces, each with its own set of obstacles impacting detection accuracy. Researchers often address the different processing steps, static object recognition, border segmentation, and relinquishment checking, in separate studies because of the complexity of these systems. There has not been much published on how each step improves the overall output of the pipeline as progress is rarely applied to a whole pipeline though it is conceived at stage. One the most important duties of surveillance systems in defending the public, is the provision of the function to recognize abandoned objects. This research investigates the possibility of using lightweight Deep Convolution Neural Networks (DCNN) for abandoned item recognition in surveillance environments in real- time. The purpose of abandoned item detection is to identify things that have been overlooked for a long time and could be a security risk or a logistical challenge. Traditional approaches often lack greatly in terms of precision and efficiency. Lightweight DCNN systems are offered by compromising between computing economy and performance. With a focus on the integration of lightweight DCNN models, this paper gives an overview of recent developments and directions in abandoned object detection. Topics covered include dataset preparation, network design, training methods, and evaluation criteria. In addition, challenges and potential future work directions are highlighted.