Abstract: 2D human pose estimation, which aims to accurately estimate the position and posture of human body joints from images or videos, holds significant importance for applications such as action recognition, virtual reality, and augmented reality. This paper provides a systematic review and summary of recent research progress in human pose estimation within the field of computer vision, including discussions on methods for both single-person and multi-person pose estimation. It also introduces the advantages of lightweight multi-person pose estimation in addressing complex scenarios, discusses public datasets, performance evaluation metrics, and experimental analyses on these datasets, offering readers a comprehensive framework for technical assessment. Finally, the paper outlines the challenges facing the field of human pose estimation and proposes future research directions, aiming to provide more accurate and reliable technical support for human behavior analysis and related domains.