Action detection is a rapidly growing field that involves identifying and tracking particular body key points in photos/videos to ascertain an individual’s action. Its difficulties arise from the natural fluctuations in lighting, a wide range of clothes, and the multiplicity of body actions that are seen in videos. Robust classification algorithms that can precisely anticipate pose angles and distances are necessary to tackle these problems. Action detection has many uses in several fields. The accurate analysis of body positions in fitness tracking helps create customized workout plans. In the medical field, awareness and detection of actions is essential for patient mobility monitoring and rehabilitation. Accurate action identification is beneficial to motion analysis, enabling advances in surveillance systems, sports analytics, and animation. The difficulty of detecting actions in videos is becoming increasingly popular due to its many uses in fields such as motion analysis, fitness tracking, and healthcare. This study first, compares the performance of several classifiers, including Support Vector Machine (SVM), k-nearest neighbors, logistic regression, decision trees, naive Bayes classification, and convolutional Neural Network (CNN) in the complex problems of posture detection, and secondly estimates various actions in real-time where the model is trained on video datasets after selecting the most suited machine learning model, The main motivation behind this research is to find the most accurate and effective classifier possible and using this information to classify actions while taking input from machine’s camera in real-time.