In recent years, damage to fisheries caused by Phalacrocorax carbo occurs throughout Japan. Our research group have been developing system for driving away verminous birds by using drone. In previous research, a method that PSPNet and original classifier is proposed. But there are problems such as the inability to cope with an increase in number of target objects. Therefore, in this paper, confirmation that whether PSPNet can classify with high accuracy and discuss classification accuracy for texture. As a result, it could be confirmed that classifying with high accuracy in 4 classes airplane, car, crow, and boar.
Deep Learning
Pyramid Scene Parsing Network
Semantic Segmentation