Underwater Vision

Underwater vision research aims to overcome the challenges of poor visibility and light distortion inherent in aquatic environments to enable robust computer vision for autonomous underwater vehicles (AUVs) and other applications. Current research focuses on developing computationally efficient image enhancement techniques, often employing convolutional neural networks (CNNs), generative adversarial networks (GANs), and diffusion models, to improve image quality and enable reliable feature extraction from sonar and optical imagery. These advancements are crucial for improving AUV navigation, object detection, 3D mapping, and other tasks, ultimately furthering our understanding of underwater ecosystems and enabling more effective subsea operations.