These last few years have seen major advancements in the area of image and video analysis brought about by machine learning. These advancements have made it possible to construct models that are both highly accurate and efficient for a broad variety of applications. The purpose of this study is to present a complete review of these improvements, with a particular emphasis on important approaches such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), and generative adversarial networks (GANs). We investigate its applicability in a variety of tasks, including image classification, object identification, semantic segmentation, and action recognition in films. Even though many breakthroughs have been made, there are still a number of issues that need to be addressed. Many challenges include data quality and annotation, computational complexity, resilience to fluctuations, and ethical considerations. In addition, we talk about the most recent developments in hardware acceleration, self-supervised learning, and explainable artificial intelligence. The purpose of this article is to give insights into the present status of machine learning in image and video analysis as well as future directions by emphasising both the successes and the continuing obstacles that are currently being faced.