Artificial Underwater Dataset: Generating Custom Images using Deep Learning Models

Polymenis, I; Haroutunian, M; Norman, R; Trodden, D

doi:10.1115/OMAE2022-79891

Artificial Underwater Dataset: Generating Custom Images using Deep Learning Models

Lookup NU author(s): Ioannis Polymenis, Dr Maryam Haroutunian ORCiD, Dr Rosemary Norman ORCiD, Dr David Trodden

Downloads

Accepted version [.pdf]

Licence

This is the authors' accepted manuscript of a conference proceedings (inc. abstract) that has been published in its final definitive form by ASME, 2022.

For re-use rights please refer to the publisher's terms and conditions.

Abstract

The rapid developments in underwater operations by the offshore industry and the scientific community have led to moresophisticated Underwater Vehicles. Notably, many underwater operations such as inspection of underwater structures are done with the help of Autonomous Underwater Vehicles (AUVs). In recent years autonomous vehicles have benefited immensely from the advancements in the area of Artificial Intelligence (AI) and particularly in Deep Learning (DL). Although DL models and applications are extensively used in different areas, such as in Aerial Unmanned Vehicles, Autonomous Car Navigation, and other applications, they are not so widespread in underwater applications mainly because of the difficulties in obtaining underwater datasets for a given application. In this regard, the present work takes the recent advances that have been made in the DL field and creates a custom dataset by using images of objects taken in a controlled, lab environment. Generative Adversarial Networks (GANs) were used to generate underwater images by combining the images taken in a lab and images with underwater environment characteristics. The results showed that it is possible to create such a dataset since the generated images were close to the actual underwater conditions. Consequently, artificial datasets of the underwater environment can overcome the difficulties arising from the limited access to real underwater images and be employed in projects with specific needs in autonomous underwater operations.