Traditional Dai Ethnic Residential Architecture Classification Based on Inception V3 Model

Abstract The accurate classification of traditional Dai ethnic residential architectures is beneficial for the elucidation of the laws of homologous differentiation of Dai ethnic groups, and is of great significance for further research on the landscape division of Dai ethnic traditional settlements. However, at present, Nevertheless, the creation of a dataset of traditional Dai ethnic residential architectures has yet to be undertaken, and no literature on the deep learning algorithm for the classification of these structures could be retrieved. To address this gap in the existing literature , this paper presents a collection of image data on traditional Dai ethnic residential archi-tectures in the Pu'er region of Yunnan Province, China. The data was collected through field research. The original data have been subjected to cleansing procedures in order to construct a da-taset comprising five distinct types of traditional Dai ethnic residential architectures. The five types of traditional Dai ethnic residential architecture are as follows: 'fence-standing with square roof', 'fence-standing with rectangular roof', 'courtyard-standing with double-eaved sloped roof', 'courtyard-standing with single-eaved sloped roof', and 'traditional earthen flat-roof house'. Following the data enhancement, the Inception V3 network model was employed for classification. The model demonstrated superior performance, achieving a classification accuracy of 93.41%, which surpassed the accuracy, recall, and loss rate of ResNet50 (Residual Network 50, ResNet50) and AlexNet(Alex Krizhevsky Network, AlexNet). This artificial intelligence classification method offers an efficient and intelligent method for elucidating the genetic characteristics and landscape differences of different types of Dai ethnic traditional residential architectures. This approach facilitates the transition from manual to intelligent genealogical study of traditional residential architectures.