Human Action Detection using EfficientNetB3 Model
Kanwarpartap Singh Gill, Avinash Sharma, Vatsala Anand, Kulbhushan Sharma, Rupesh Gupta · 2023
A result of the swift developments in computer vision and machine learning, careers in video analysis have been transitioning from inferring the current state to forecasting the future state. These include the tasks of vision-based action prediction and recognition with social security in public spaces or amongst the population, where action prediction predicts human actions based on partially accomplished activities and action recognition predicts human actions based on fully executed actions with precision. These two challenges have recently gained a lot of attention due to their real-world applications, which are expanding quickly in fields including video retrieval, autonomous driving, entertainment, and visual surveillance. As a result, there have been many efforts over the past few decades to provide a solid and effective framework for action recognition and prediction. In this research, the proposed model shows the accuracy of value 76 percent on the Adam optimizer using EfficientNetB3 model.