Deep Learning it has been the subset assortment of Machine Learning concerned where neural system calculations enlivened by human cerebrum (what happens immediately to human) gain from enormous measure of information through a few layers for nonlinear change. The deep learning can process huge number of highlights to build the result exactness. Genuine applications on Deep Learning, Face Recognition, Hand Writing Recognition, Speech Recognition, translate starting with one human language then onto the next human language, Control Robots such as self-driving vehicles. The current framework depends on sensors and gadgets to gather time arrangement signals which are created in both time and recurrence space. To accumulate the stimulating information, each subject conveys a keen gadget for a couple of hours and plays a few exercises. In the anticipated application, five sorts of basic exercises will be actualized, including strolling, limping, working out, strolling upstairs, and strolling downstairs. Human Activity Recognition (HAR) has expanded a lot in look into field especially setting mindful figuring and sight and sound-generally on the record of its pervasiveness in human life and besides on our reliably growing computational limit. It is generally speaking adequately looked for after for a wide scope of employments like sharp homes, human direct examination, sports and even security systems. The proposed application Human Activity Recognition depends on Deep Learning which is utilized to recognize and check the human exercises from the pictures. Deep Learning Algorithms influence enormous datasets of old human exercises and gain from rich arrangement of highlights and train the models and in the long run beat the human exercises. The proposed application included Feature Detection, Feature Alignment, Feature Extraction, Feature Detection.