Research on Underwater Image Recognition based on Transfer Learning
Jiajia Zhou, Junbin Zhuang, Benyin Li, Liang Zhou · OCEANS 2022, Hampton Roads · 2022
To address the problem that underwater image data is difficult to obtain and the quantity is small, this paper proposes an image recognition method based on improved secondary transfer learning on the basis of direct transfer learning. Firstly, cifar-10 is used as an intermediate transition data set for secondary transfer learning to improve the accuracy of underwater image recognition network model. Secondly, the fully connected layer of the constitutional neural network is designed to be refined and simplified for the problems of many parameters, slow training and large storage capacity of the secondary transfer network. The experimental results show that the accuracy of the streamlined model is improved by 3.33% compared with direct transfer learning.