A Comparison of Few-Shot Learning Methods for Underwater Optical and Sonar Image Classification

摘要: Deep convolutional neural networks generally perform well in underwater object recognition tasks on both optical and sonar images. Many such methods require hundreds, if not thousands, of images per class to generalize unseen examples. However, obtaining labeling sufficiently large volumes data can be relatively costly time-consuming, especially when observing rare objects or performing real-time operations. Few-Shot Learning (FSL) efforts have produced many promising deal with low availability. little attention has been given the domain, where style poses additional challenges for algorithms. To best our knowledge, this is first paper evaluate compare several supervised semi-supervised using side-scan imagery. Our results show that FSL offer a significant advantage over traditional transfer learning fine-tune pre-trained models. We hope work will help apply autonomous systems expand their capabilities.