The importance of crowd control and supervision in ensuring public safety is widely recognized, and this study area continues to evolve. However, developing a reliable crowd monitoring system (CMS) presents several challenges, such as uneven object distribution, varying densities, occlusions, and difficulties with pose estimation. CCTV cameras are commonly used in hospitals, parks, stadiums, airports, and cultural or religious sites to monitor crowds. Despite their utility, these cameras have several drawbacks, including limited coverage, installation challenges, mobility issues, high power consumption, and the necessity for continuous human oversight. Many researchers are exploring computer vision and machine learning solutions to address these issues and reduce the need for human intervention. This article reviews and categorizes recent research on crowd monitoring that employs various machine learning algorithms and techniques published in journals and conferences.