Human Action Recognition Using Deep Learning Methods (CNN-LSTM) Without Sensors

Surveillance System is gaining attention in recent years in the computer vision field that's where the Human Activity Recognition using videos feed plays an important role. The Human Activity Recognition is a Challenging task to implement in real life technology because of the high computation and accuracy involved. This paper proposes a method for Recognising Activities from video input feed. In this paper we are using Long Short Term Memory (.LSTM) and Convolutional Neural Network (C.N.N) to classify the videos and train the model. This model is trained using UCF-50 Dataset, it consists of 50 activities. The HAR system is widely used in many applications such as Surveillance, healthcare, security fields etc. The main objective of this paper is to recognize activities performed by the humans using videos dataset as an input and using machine and deep learning libraries.