Gender recognition of Guanyin in China based on VGGNet

Research reveals that the gender transformation of Guanyin (Avalokitesvara) in China is more influenced by geographical factors than temporal ones, challenging traditional historical narratives. Using VGGNet, a deep convolutional network, the study objectively quantified gender characteristics in Buddhist images from Dazu Rock Carvings, Dunhuang Mogao Caves, and Yungang Grottoes. Findings indicate that Guanyin's gender was regionally consistent rather than evolving uniformly over time, with prominent male traits in Dunhuang frescoes and feminine features in earlier Yungang carvings. The study also highlights the importance of culturally appropriate training datasets for accurate gender recognition, demonstrating that Indian facial data is less effective for Chinese Buddhist imagery due to sinicization.

Huang, Y., Chen, D., Wang, H., & Wang, L. (2022). Gender recognition of Guanyin in China based on VGGNet. Heritage Science, 10(1). https://doi.org/10.1186/s40494-022-00732-3