Human Activity Recognition Using Efficientnet-B0 Deep Learning Model

Human activity has many varieties of applications. By using HAR one can predict the behaviour of a person, one can predict the wellness of a person, whether he or she is healthy or not. It can be used in surveillance area, by which one can detect the abnormal activities of a person. Also HAR helps to detect the human activities in our daily life. In this study, implementation of efficientNet-Bo provided a model accuracy of 72 percentage for 31 epochs. This model gave a highest accuracy of 74 percentage for the epoch 11. By implementing this model on HAR dataset we can detect the human activities calling, clapping, cycling, dancing, drinking, eating, fighting, hugging, laughing etc.