Many models have been proposed in the active field of human activity recognition research and scientific studies, using different techniques for activity identification and classification with ML. In the groups of kinetic models concerning leaning on spatial or temporal features being used for capturing the features of sets of image or video. Moreover, several successfully implemented deep layer trained models in this domain have served the primary goal of the model, which is the detection and classification of the activity taking place. The types of these activities can be of numerous kinds such as mundane activities like eating, sitting, jogging, or running, etc. Activity identification has been a field of study for decades. Inherent, high requirements for it have for ambient assisted living, intelligent surveillance, human-computer interaction and many other applications. In this work we have explored the integration of CNN with a Long-Short Term Memory Network to identify human activity in films. For example, two different architectures and methods (Conv LSTM and LRCN) are built on the UCF50 activities dataset using Tensor Flow. A tip-off of suspicious activities occurring instantaneously is sent to the authority through the server according to the detected nature. A lot of bad things can be avoided, or at least their bad effects can be minimized, by using this paradigm.