A Fast Sonar Image Style Transfer Model Based on Improve Muti-Style-Per-Model
Xie Zhi-min, Zhou Xilin, Yanhui Wei, Hou Jianing · 2024
Sonar images are an important way and carrier of underwater information, playing a crucial role in the exploration, development, and utilization of marine resources. However, there are few publicly available datasets for sonar images. With the rapid development of deep learning technology, there are more and more mature style transfer models based on optical images. This article briefly outlines various style transfer networks, improves them based on fast style transfer learning algorithms, and proposes a style transfer algorithm for sonar images. The collected optical images are used to generate sonar data through the transfer algorithm, providing a dataset for subsequent experiments. Compared with other style transfer models, our model has shown better performance in images after transfer under two commonly used evaluation metrics.