Multi-task Learning for Identification of Porcelain in Song and Yuan Dynasties

2025-03-18

AI Summary

To address the low efficiency, high subjectivity, and poor scalability of manual authentication of Song-Yuan dynasty ceramics, this paper proposes an end-to-end deep learning classification framework for multi-attribute recognition (dynasty, glaze color, kiln site, and vessel form). We introduce, for the first time, a multi-task learning architecture integrated with lightweight CNNs (MobileNetV2 and ResNet50) for fine-grained, cross-attribute ceramic classification. We systematically evaluate the substantial performance gains of transfer learning in small-sample cultural relic classification and design domain-specific pretraining strategies alongside interpretability-enhancement mechanisms. Experiments demonstrate that MobileNetV2 and ResNet50 achieve an average accuracy exceeding 92% across the four joint tasks; notably, vessel-form classification improves by up to 18.5% over ablated baselines trained from scratch. These results validate the proposed method’s effectiveness, robustness, and generalizability for fine-grained cultural heritage classification.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Representation Learning for Vision
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Chinese porcelain holds immense historical and cultural value, making its accurate classification essential for archaeological research and cultural heritage preservation. Traditional classification methods rely heavily on expert analysis, which is time-consuming, subjective, and difficult to scale. This paper explores the application of DL and transfer learning techniques to automate the classification of porcelain artifacts across four key attributes: dynasty, glaze, ware, and type. We evaluate four Convolutional Neural Networks (CNNs) - ResNet50, MobileNetV2, VGG16, and InceptionV3 - comparing their performance with and without pre-trained weights. Our results demonstrate that transfer learning significantly enhances classification accuracy, particularly for complex tasks like type classification, where models trained from scratch exhibit lower performance. MobileNetV2 and ResNet50 consistently achieve high accuracy and robustness across all tasks, while VGG16 struggles with more diverse classifications. We further discuss the impact of dataset limitations and propose future directions, including domain-specific pre-training, integration of attention mechanisms, explainable AI methods, and generalization to other cultural artifacts.

Problem

• Automate classification of Song and Yuan porcelain artifacts.
• Enhance accuracy using deep learning and transfer learning.
• Address dataset limitations and propose future AI improvements.

Innovation

• Uses multi-task learning for porcelain classification
• Applies transfer learning with CNNs for accuracy
• Evaluates four CNN models for performance comparison