Virtual Underwater Datasets for Autonomous Inspections

Underwater Vehicles have become more sophisticated, driven by the off-shore sector and scientific community's rapid advancements in underwater operations. Notably, many tasks, including assessment of subsea infrastructure, are performed with assistance Autonomous (AUVs). There been recent breakthroughs Artificial Intelligence (AI) and, notably, Deep Learning (DL) models applications, which widespread usage a variety fields, aerial unmanned vehicles, autonomous car navigation, other applications. However, they not as prevalent applications due to difficulty obtaining datasets for specific application. In this sense, current study utilises area DL construct bespoke dataset generated from photographs items captured laboratory environment. Generative Adversarial Networks (GANs) were utilised translate object into domain combining collected images containing The findings demonstrated feasibility creating such dataset, since resulting closely resembled real environment when compared real-world ship hull images. Therefore, artificial can overcome difficulties arising limited access used enhance operations through image classification detection.