Advancements and challenges in computer vision: From pixels to perception

Computer vision has undergone remarkable transformation during the last decade, driven largely by advances in deep learning, foundation models, generative artificial intelligence, and three-dimensional vision technologies. These developments have improved performance in many areas, including healthcare, autonomous driving, robotics, manufacturing, and security. Despite these achievements, high accuracy alone does not guarantee reliable performance in real-world environments. This paper reviews the major developments and challenges in modern computer vision. The discussion focuses on deep learning architectures, transfer learning, three-dimensional vision, generative models, data quality, explainability, robustness, and ethical issues. It highlights the need for computer vision systems that are not only accurate but also transparent, reliable, and trustworthy. In addition, a conceptual framework is presented to illustrate how these challenges interact across different stages of the computer vision pipeline.