Augmented Reality for Industrial Maintenance Using Deep Learning Techniques– A Review

Industry 4.0 represents the forefront of the ongoing industrial revolution, wherein the objective is to intricately merge the realms of digital and physical domains within industrial processes, with the ultimate aim of enhancing the efficacy of catering to the diverse requirements of various stakeholders. This study presents a pioneering prototype of Pervasive Augmented Reality (AR), which has been devised in conjunction with esteemed industrial collaborators. The primary objective of this prototype is to facilitate the instantaneous monitoring of data and the identification of issues within an industrial assembly line. The primary emphasis lies within the preprocessing procedures, semantic segmentation and object detection which enable the model to manifest as a virtual scene for various maintenance-related endeavors, such as off-site maintenance scheduling or revamping. A Human-Centered Design (HCD) methodology was employed to ascertain the intricacies, requisites, and predicaments faced by stakeholders, while concurrently establishing the prerequisites to steer the evolution of the model. The initial observations of these individuals are delineated, subsequent to an inaugural user study conducted within a factory shop floor setting, aimed at assessing the proposed prototype and gathering input on potential enhancements to optimize its efficacy within such contextual domains.