Research and Application of Ancient Chinese Pattern Restoration Based on Deep Convolutional Neural Network.

Area of Science:

• Artificial Intelligence
• Computer Vision
• Machine Learning

Background:

• Deep learning, a subset of machine learning derived from artificial neural networks, excels in image recognition and data processing.
• Deep convolutional neural networks (CNNs) utilize multilayer networks to extract features from images, enabling identification and classification.
• The objective of deep learning is to replicate the analytical and learning capabilities of the human brain in machines.

Purpose of the Study:

• To comprehensively review the application and research progress of deep convolutional neural networks in ancient Chinese pattern restoration.
• To introduce the fundamental knowledge of deep convolutional neural networks and summarize related algorithms.
• To analyze the key structures and principles of traditional pattern repair methods based on deep convolutional neural networks.

Main Methods:

• Developing image restoration models utilizing deep convolutional networks and adversarial neural networks.
• Implementing a model comprising information masking, feature extraction, a generative network, and a discriminative network.
• Testing the deep convolutional neural network-based method against traditional and other deep learning restoration techniques on Qinghai traditional embroidery datasets.

Main Results:

• The proposed deep convolutional neural network method demonstrated superior evaluation indices compared to traditional sample-based and other deep learning restoration methods.
• Visual inspection revealed that the deep convolutional neural network method produced more aesthetically pleasing and human-eye-aligned restoration results.
• The developed model's components (information masking, feature extraction, generative, and discriminative networks) function both independently and interdependently for effective restoration.

Conclusions:

• Deep convolutional neural networks offer a powerful and effective approach for the restoration of ancient Chinese patterns.
• The developed deep convolutional neural network and adversarial network model significantly improves upon existing image restoration techniques.
• The method's success in accurately and visually appealingly restoring intricate patterns highlights its potential for cultural heritage preservation.