The article Human Activity Recognition from Video Captured Sequences is used to build a model that would take video input from multiple sources, including security cameras and recorded movies, and identify actions that are being done in them by humans. The term “human activity recognition” or “human activity detection” is another name for this procedure. Research on video surveillance is a hot topic that includes classifying human behavior into ordinary (normal), unusual (abnormal), and suspicious actions. Additionally, the classification of human actions is done using the CNN algorithm in two different ways: rapid classification of video frames and complete classification of video sequences. The Human Motion Database—HMDB can be used to evaluate human detection and classification of human activity. Our HMDB dataset, which consists of 7,821 video sequence images, is used for the purpose of training our human activity recognition system. After 20 epochs, the method achieved an effective recognition accuracy of 98% with 22 frames/second.