An adaboost-backpropagation neural network for automated image sentiment classification

Abstract

The development of multimedia technology and the popularisation of image capture devices have resulted in the rapid growth of digital images. The reliance on advanced technology to extract and automatically classify the emotional semantics implicit in images has become a critical problem. We proposed an emotional semantic classification method for images based on the Adaboost-backpropagation (BP) neural network, using natural scenery images as examples. We described image emotions using the Ortony, Clore, and Collins emotion model and constructed a strong classifier by integrating 15 outputs of a BP neural network based on the Adaboost algorithm. The objective of the study was to improve the efficiency of emotional image classification. Using 600 natural scenery images downloaded from the Baidu photo channel to train and test the model, our experiments achieved results superior to the results obtained using the BP neural network method. The accuracy rate increased by approximately 15% compared with the method previously reported in the literature. The proposed method provides a foundation for the development of additional automatic sentiment image classification methods and demonstrates practical value.