Work

Chronological classification of Ming and Qing dynasty ceramics images based on an enhanced ResNet50 model

Year: 2025

Type: article

Abstract: To overcome the limitations of expert-dependent and conventional methods in Ming-Qing ceramic chronology analysis, we propose an enhanced ResNet50 model that integrates dual attention mechanisms (channel and self-attention) to amplify discriminative visual features. A reliable dataset of Ming and Qing ceramics was constructed for training and evaluation. Experimental results show that the improved “Ours” model achieves 94.75% accuracy on the test set, a 2.88% improvement over ResNet50, and outperforms advanced models such as VGG16, DenseNet121, and MobileNetV2. Cross-dataset validation indicates that the method shows excellent generalizability for both ceramic type classification and the chronological categorization of other archaeological artifacts. These findings demonstrate that the integration of both channel and self-attention mechanisms effectively resolves the visual feature differences in Ming and Qing ceramics, providing reliable technical support for the digital preservation of cultural heritage. The source code is publicly available at https://github.com/ZhouAlone/ceramic/.

Source: STAR Science & Technology of Archaeological Research

DOI: 10.1080/20548923.2025.2498260

Authors: LiHua Yang, WenBo Zhou, WangRen Qiu

Language: English

Topic: Image Processing and 3D Reconstruction

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: gold

Estimated APC paid: $969

Funders: Education Department of Jiangxi Province, Jingdezhen Science and Technology Bureau

Award: G8372160036