Comparative Evaluation of MediaPipe and YOLOv8 for Real-Time Pose Estimation

Pose estimation has emerged as a critical task in computer vision, driving advancements in applications ranging from human-computer interaction to sports analytics. This study presents a comparative evaluation of two state-of-the-art pose estimation models, MediaPipe and YOLOv8, assessing their performance under various configurations (lite, full, heavy for MediaPipe; nano, small, medium, large, extra-large for YOLOv8) across different video conditions. The evaluation metrics include average frames per second (FPS), CPU usage, and memory consumption, tested on scenarios involving walking, break dance, martial techniques, and crowded scenes. Our results demonstrate that MediaPipe’s lite configuration consistently delivers high FPS and low latency, making it suitable for real-time applications, whereas YOLOv8 excels in complex scenes such as crowded environments, showing superior handling of occlusions and dense object detection. This comprehensive analysis provides valuable insights for selecting appropriate pose estimation models tailored to specific application needs, highlighting the strengths and trade-offs of each approach.