This project aims to develop a deep learning based computer vision system to automatically detect and classify biofouling intensity on submerged infrastructure such as ship hulls, offshore platforms, and underwater pipelines. Biofouling is the accumulation of marine lifeforms on submerged surfaces, and poses significant economic and environmental challenges, including corrosion, increased fuel consumption due to drag, infrastructure maintenance cost, and spread of invasive species.
Our goal is to build a robust real-time image-based model that classifies fouling level (eg. none/low/moderate/heavy). This model will be developed using pre-trained convolutional neural network architecture (eg. YOLO and other transfer learning models) and automatically supplementing against an expert assessment to improve prediction. We will assess reliability and practical deployment feasibility of our underwater object detection model, and if possible explore extensions of our classifier model to detect