Deep Learning Based Deep-Sea Image Enhancement and Animal Species Classification

Marine research and biodiversity conservation relies crucially on underwater image analysis. Despite being an important asset for oceanographic work, the effective use of underwater imagery for various tasks remains inhibited by challenges like poor visibility, low contrast and color distortion. In this paper, I introduce a novel framework that uses FunieGAN for underwater image enhancement followed by Yolov8 for sea creature detection and classification. To improve the visual quality of underwater images the proposed approach makes use of FunieGAN's capabilities aimed to combat the haze and color inconsistency problems. Extra images are then generated using Yolov8, which are then used to detect and classify marine species with high accuracy and precision. The framework also includes a real time deployment component including a Streamlit based application to process time series underwater images in an accessible and efficient manner. The combined methodology yields significant improvements in image quality and detection accuracy that have been validated through results. This work presents a scalable and practical approach for underwater image analysis that has applications in marine biology, environmental monitoring, and fisheries management.