Research On Fish Classification And Recognition Based On Improved VGG-16 Model

This study aims to utilize deep learning technology for fish identification and detection, implementing a series of improvements to the limitations of the traditional VGG-16 network model in order to enhance the model’s detection capabilities and recognition accuracy, as well as providing technical support for fish conservation and research. First, the data was fully preprocessed, including image enhancement and normalization, to improve data quality. Next, batch normalization was used to modify the VGG-16 network structure in order to improve the model’s stability and generalizability. The improved model achieved an accuracy of $98.89 \%$ on the test set, which is a significant improvement over the traditional VGG-16 model. We conducted an in-depth analysis of the model, explored the specific contributions of various optimization measures to performance improvement, and verified its feasibility and effectiveness through experimental results. The results show that data preprocessing and network structure adjustment are the key factors for performance improvement. The results of this study show that the improved deep learning model has high accuracy and reliability in fish identification and detection. This study is of great significance for promoting the application of deep learning technology in the fields of biodiversity conservation and fishery management, in parallel with providing an effective approach and reference for optimizing deep learning models.