Related references
Note: Only part of the references are listed.
Article
Chemistry, Analytical
Lucas C. F. Domingos et al.
Summary: This paper provides a comprehensive overview of current deep-learning methods for automatic object classification of underwater sonar data. It identifies five main ingredients in the development of this field and discusses important datasets and data-augmentation techniques.
Article
Automation & Control Systems
Yannik Steiniger et al.
Summary: This paper provides a comprehensive overview of the application of deep learning in the analysis of sonar images. The focus is on feature extraction, classification, detection, and segmentation using convolutional neural networks (CNN). The research in this field has shown that even small CNNs outperform traditional methods. The purpose of this work is to introduce researchers to the recent achievements in this field and to identify research gaps for further exploration.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2022)
Article
Engineering, Civil
David P. Williams
Summary: By using tiny CNNs with significantly fewer parameters, this study demonstrates classification performance that matches or even surpasses that of human domain experts. The research represents the first large-scale classification study in the sonar domain, showing that CNNs have strong generalization ability in challenging environments, which will significantly impact their utilization in the underwater remote-sensing community.
IEEE JOURNAL OF OCEANIC ENGINEERING
(2021)
Article
Engineering, Marine
Yannik Steiniger et al.
Summary: The article presents a method of using GAN to generate synthetic sidescan snippets. Training GAN with few data samples may cause mode collapse, but pre-training with simulated images and fine-tuning with real data can alleviate this issue. However, this approach is sensitive to the pre-training step when applied to sonar data due to the vanishing of the gradients of the GAN's discriminator.
JOURNAL OF MARINE SCIENCE AND ENGINEERING
(2021)
Article
Computer Science, Information Systems
Xiaoming Qin et al.
Summary: Acoustic seabed classification combined with deep learning was utilized to study sediment classification in shallow-water, side-scan sonar images from the Pearl River Estuary. Applying different and deeper CNNs with pretraining on grayscale CIFAR-10 achieved parameter migration and improved model performance, with ResNet reaching a 3.459% error rate after fine-tuning. Data enhancement through GANs showed potential for improving sediment classification accuracy, but with limited effects and computational expenses.
Review
Computer Science, Information Systems
Dhiraj Neupane et al.
Proceedings Paper
Engineering, Marine
Marija Jegorova et al.
GLOBAL OCEANS 2020: SINGAPORE - U.S. GULF COAST
(2020)
Proceedings Paper
Computer Science, Artificial Intelligence
Kaiming He et al.
2017 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV)
(2017)
Article
Geochemistry & Geophysics
Tai Fei et al.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2015)