Enhancing Human Action Recognition in Diverse Environments using HADE Dataset

Abstract

Human Action Recognition (HAR) is an important aspect of Computer Vision, because it enables the identification and classification of human actions using various sensors and cameras. This technology has extensive applications across sectors, such as security, sports, healthcare, and entertainment. Deep neural networks have the capability to manage the complex behavior of HAR systems by capturing complex spatiotemporal features, leading to dynamic action recognition and promoting the performance of HAR systems to unique levels. However, achieving full potential depends on access to high-quality, diverse datasets and advanced machine learning algorithms. This research created a new dataset called Human Actions in Diverse Environments (HADE), which focused on four critical human actions. The aim was to improve the effectiveness of human action recognition across multiple domains as well as to develop practical applications for this technology in the real world. This research draws attention to the interrelated factors that contribute to the transformative capability of HAR, emphasizing its advancement toward a wider scope of Computer Vision.