Hybrid Framework for Underwater Coral Image Recognition Using Generative Adversarial Networks and Semi-Supervised Segmentation

Underwater environments are highly complex, where light attenuation, color shifts, and suspended particles degrade image quality and hinder reliable coral recognition. To address this, we propose an underwater coral image recognition framework that combines generative adversarial networks with semi-supervised segmentation. In this framework, the generative model enhances low-quality images and reduces domain discrepancies, while the semi-supervised strategy enables accurate semantic segmentation and species recognition with limited labeled data. Experiments on publicly available underwater coral datasets show that the proposed model outperforms traditional methods in terms of mean Intersection over Union (mIoU) and recognition accuracy. These results demonstrate a promising approach for underwater ecological monitoring and coral reef conservation.