In image processing and computer vision, human activity detection is a significant activity. There are various techniques and approaches for key point detection that identify the external Skelton key points. Some methods will detect key points and recognize the human pose. The proposed work aims to utilize Random Forest (RF) approach and classify the human activity into 15 classes using media pipes. The library trained with 30,000 samples. The objective of this paper is to capture the human face, like the angles of limbs and key points and train the machine-learning model to recognize the human action using media pipe. In future, we can extend this work to capture real-time video poses using intelligent methods for key points to identify the actions of human facial expressions.