Integrating Attention Mechanisms into YOLOv8 for Improved Defect Detection in Acceptance Environments on HoloLens 2

The advent of augmented reality (AR) devices, such as HoloLens 2, has paved the way for interactive and immersive inspection experiences. This paper delves into the design and implementation of a scalable mixed reality system, tailored to support contactless defect detection and provide real-time feedback to acceptance personnel in industrial settings, particularly within the context of power substation acceptance. Our approach integrates sophisticated neural network architectures, enhanced with attention mechanisms, into the YOLOv8 framework. This integration is specifically aimed at bolstering detection precision and processing speed on HMDs like HoloLens 2. By harnessing the power of HoloCV—a framework that streamlines computer vision tasks on HoloLens—our system is capable of detecting and visually representing defects in real time. This includes critical issues such as honeycombing in concrete foundations and paint peeling on metal components, which are common in substation environments. The system's ability to provide real-time, visually enhanced feedback on defects is a significant advancement for acceptance scenarios, particularly in high-stakes settings like power substations. It allows for immediate identification and assessment of potential hazards, thereby facilitating swift corrective actions. This not only streamlines the inspection process but also ensures a higher standard of quality control, which is paramount for maintaining the safety and reliability of electrical infrastructure. The integration of advanced AR technology with cutting-edge defect detection algorithms offers a transformative approach to substation acceptance. It empowers industry professionals with the tools necessary to uphold rigorous quality standards, ensuring that electrical facilities meet the highest levels of operational integrity and safety compliance.