Action Recognition Using Segmental Action Network

Abstract

Human action recognition refers to the task of recognizing and categorizing human actions in video or image sequences.This is a complex problem in computer vision and has a wide range of applications, including video surveillance, human-computer interaction, and sports analysis.In this work, A novel approach to action recognition using Segmental Action Networks is presented.The proposed approach utilizes 2D and 3D convolutional neural networks to extract spatiotemporal features from video frames, which are used to train a Segmental Action Network.To improve the model's accuracy, Different voting and feature extraction techniques, such as Space-Time Interest Points, (STIP) and Optical flow have been applied.The proposed model has been tested on the HMDB51 dataset and has achieved better results than existing models.The results demonstrate the effectiveness and robustness of our proposed approach for action recognition.Furthermore, our model is computationally efficient and can be deployed on edge devices with low computational and memory capacity, making it a promising approach for real-world applications.