Transfer Learning with deep Convolutional Neural Network for Underwater Live Fish Recognition

International audience; Recently, marine ecologists are more interested in using underwater video analysis to study fish populations as this technique is non-destructively, produces large amount of visual data and does not affect fish behavior. Automated methods for processing and analyzing the recorded data are required because visual analysis can be subjective, time consuming and costly. However, the underwater environment poses great challenges due to changes in luminosity, complex backgrounds and free movement of fish. In this paper, we present a convolutional neural network that was trained with transfer learning framework for fish species classification. First, we extract fish features from images using the original AlexNet model on the available underwater dataset. Then, to improve the performance, we fine-tune the model on the dataset. Finally, we re-extract features after that AlexNet has been fine-tuned. For classification, we use a linear SVM classifier. The experiment results demonstrate the effectiveness of the proposed approach, we achieve an accuracy of more than 99%.