A Comparison Study of Abnormal Human Activity Analysis

Recent years have seen a substantial increase in interest to researchers in the modelling of human behaviors and activity patterns for the detection of abnormal events. Many approaches, including the implementation of monitoring systems, have been undertaken to stop the abnormal human actions. If the monitoring systems can detect unusual human activity automatically and send out alarm or warning signals, that will be quite significant. The first step is for the algorithm to recognize whether there are any people in a frame of footage. Then, it's necessary to remove the frames that are likely to include abnormal human behavior. At this time, the useless frames should be removed. When a human exhibits abnormal behavior, the trained model identifies it and distinct photos of those frames are kept. The ability to identify faces in these pictures has been improved. Here is the requirement to develop an automated security system that identifies the abnormal human activity in real-time so one can immediately take action on it. It is a very lengthy process to get abnormal human activity from lengthy surveillance videos so it will compress the video before passing it throw the activity recognition system so that system can first retain the objects of interest and then it can be passed throw the model. Utilizing just CNN (Convolutional Neural Network) is less accurate and consumes a lot of computing time. As a result, MobileNet, a pre-trained model, is used as the foundation for developing the complete model and offers improved accuracy. Telebot uses the Telegram app to send an alarm message to the relevant authorities.