Facial Recognition and Classification of Terracotta Warriors in the Mausoleum of the First Emperor Using Deep Learning

A deep learning model achieved 95.6% accuracy in recognizing and classifying the facial features of Terracotta Warriors from the Mausoleum of the First Qin Emperor, surpassing classical models like SqueezeNet and Inception_ResNetV1 by 4.1% and 2.8%, respectively. The study enhanced the FaceNet backbone feature extraction network with three 3×3 convolution kernels instead of a 7×7 kernel, incorporating Fire modules and an exponential loss function for smoother training. Agglomerative Clustering was then used to classify 295 warriors, providing a robust technical foundation for technological archaeology. This approach offers efficient and intelligent support for analyzing the authentic depictions of soldiers from the Qin Dynasty.

Sheng, Y. (2024). Facial Recognition and Classification of Terracotta Warriors in the Mausoleum of the First Emperor Using Deep Learning. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, X-1-2024, 205–212. https://doi.org/10.5194/isprs-annals-x-1-2024-205-2024