The Role of Artificial Intelligence and Machine Learning in Computer Vision: From Feature Engineering to Generative Synthesis

Abstract

The domain that has completely changed after the introduction of Artificial Intelligence (AI) and Machine Learning (ML) is Computer Vision (CV), the sphere that allows machines to process and extract meaningful data out of the digital image and video. In this paper, the author conducts a thorough overview of the development of the AI/ML paradigms as they have progressed over augmenting the classical feature-based approaches to becoming the fundamental driving force of the current system of CV. We follow the history down to starting with the early classifiers, such as Support Vector Machines (SVM) using manually-constructed features to classifiers powered by deep learning ushered in by Convolutional Neural Networks (CNNs), capable of automatically extracting features and performing tasks such as object detection and image classification better than a human being. The review also examines the state of the frontiers, such as use of vision transformers, image synthesis and image augmentation by generative models, and multimodal learning which combines visual information with other modalities. By relying on a broad spectrum of applications, such as in the fields of medical image analysis as well as autonomous systems but also content-based retrieval and behavioral understanding, we combine results of the most relevant literature to point out breakthroughs, current challenges, as well as future trends. Such crucial problems as model interpretability, data-efficiency, algorithm bias, and real-time processing constraints are discussed in details. We conclude that AI/ML has not only improved computer vision but changed its very nature, making it possible to perform more than ever before in terms of perception, understanding, and creation of visual content with significant effects in the fields of science, industry, and society.

DOI: https://doi-ds.org/doilink/01.2026-15735329