The pervasive integration of modern artificial intelligence into daily life necessitates robust human computer interaction, underscored by advancements in computer technology. As the primary human tool, hand position, and orientation are vital for applications like virtual reality, with hand pose estimation playing a pivotal role. Despite the challenges inherent in RGB image-based estimation, the coalescence of extensive datasets, neural networks, and heightened computational capabilities has catalyzed the utilization of deep learning in visual tasks, particularly advancing hand pose estimation. This review exclusively centers on methodologies of RGB image-based approaches within the realm of hand pose estimation. The paper explores recent advancements in this category, providing a comprehensive analysis of outcomes and charting future directions. Organized around discussions on challenges, advancements, and future directions, this review aims to contribute to the ongoing discourse in the dynamic RGB image-based hand pose estimation field. By addressing unique challenges specific to this method, the paper offers insights that contribute to developing precise and robust hand pose estimation systems, impacting the broader landscape of computer vision and human-machine interaction.

Keywords