Review of machine learning-based techniques for detecting specific and general objects

In recent years, machine learning has become an essential technology across multiple areas, particularly in object detection systems, where it enhances accuracy and automation. This paper critically examines the object detection techniques developed between 2020 and 2024, focusing on three popular methods, namely You Only Look Once (YOLO), Single Shot Detector (SSD) and Convolutional Neural Network (CNN). These architectures have been assessed based on their performance in different implementation environments, highlighting strengths and limitations. Further, the architectural layers of each model are discussed to offer deep insights into their operational mechanisms. YOLO, known for its fast inference speed and adaptability, is particularly suited for real-time surveillance and autonomous navigation applications. Specialized models such as Fuzzy C-Means are also explored for their advantages in object segmentation, tracking and multi-object detection. The paper concludes by examining the future direction of object detection, emphasizing the need for more efficient and adaptable models to integrate machine learning into everyday applications.