Sex Classification of Salmon Using Convolutional Neural Network

In this study, we attempted to classify the sex of salmon using a convolutional neural network. We collected labeled(male/female) images of salmon and other fishes, and trained the VGG16 type neural network by implementing several data argumentation techniques, cropping the bounding box of salmon. In evaluation, the F value of the classification result was found to be satisfactory by more than 99%. We demonstrated the focus point of the neural network using Grad-CAM and the classification ability depending on the background processing.