Research on underwater target detection technology based on deep learning

Underwater target detection aims to locate and identify targets in underwater scenes, which plays a significant role in the development of the marine economy and environmental protection. However, existing target detection algorithms often cannot meet people’s needs in underwater environments. In this paper, we propose an underwater target classification convolutional neural network (UTC-CNN) based on You Only Look Once Version 8. This network builds a C2f_Squeeze and Excitation module to enhance the feature extraction ability of blurred underwater targets, and adopts Bidirectional Feature Pyramid Network to replace Feature Pyramid Network to further enhance the feature fusion capability. The experiments show this model outperforms other target detection models on Underwater Robot Professional Contest 2020 dataset, with mAP0.5 reaching 84.4%, demonstrating the advantages of UTC-CNN in underwater target detection tasks.