Classification and Recognition of Turtle Images Based on Convolutional Neural Network. Issue 5 (March 2020)

• Record Type: Journal Article
• Authors: Liu, Jiangchuan; Wang, Mantao; Bao, Lie; Li, Xiaofan; Sun, Jun; Ming, Yue
• Is Part Of: IOP conference series. Volume 782:Issue 5(2020)
• Publication Date: 2020-03
• Subjects: Materials science -- Periodicals; 620.1105
• DOI: 10.1088/1757-899X/782/5/052044
• Languages: English
• ISSNs: 1757-8981

Abstract

Abstract: The identification of turtle species mainly depends on the recognition of turtle head and shell, but there is no relevant study on turtle image. In this paper, a turtle image recognition and classification system based on transfer learning is introduced. The system consists of four phases. First, turtle images need to be collected for data enhancement and dataset production. Secondly, the Inception-v3 network model is used to train and save parameters and model structure on the ImageNet dataset. In addition, the network model needs to be modified to change Softmax classifier into 5 categories. Finally, the tortoise dataset was used for training and saving the model, and the classification accuracy of five representative turtles was verified. The experiment proves that the network model of migration learning adopted in this paper is faster and more accurate than the one not adopted.