Human posture estimation is important research applied in many fields such as human-machine interaction, surveillance, sports anal-ysis, etc. From there, it is possible to build intuitive and practical applications with science, technology and life. Therefore, fast and accurate estimation of human posture is a pre-processing step but very important in the process of building applications. In this paper, we propose to use Mediapipe, which is a Microsoft built-in frame-work for 3D human pose estimation. The test was evaluated against the MADS (Martial Arts, Dancing, and Sports Dataset) database, in which we focused on sports videos such as: basketball, volleyball, football, rugby, tennis and badminton. The average estimate error is between 100-200mm. The 3D human posture estimation results are a good result in supporting sports analysis.