We developed a system that was efficient enough to Detect and classify unusual activities from video streams. We worked on 3 Anomalous categories (Fighting, Shoplifting and Explosion) along with Normal and achieved accuracy of 82%. We developed a flutter application as a use case of our project that uses real time camera for video streams. In the backend our model predicts the unexpected event and sent an SMS notification containing all the information along with the user's location and current time. The model is deployed on Hureko Server.