Cultural Heritage Meets AI: Advanced Text-to-Image Models for Digital Reconstruction and Preservation
Kunta Hsieh, Tsen-Shu Tsaur, Chao Min, I-Cheng Li, Pin-Chia Huang, Minhua Lu · 2024
This paper reviews the applications and development of web crawling and text-image matching technologies in the field of information processing, with an emphasis on the CLIP model and its adaptation for Chinese contexts. Web crawling technology utilizes automated programs to simulate browser behavior, efficiently accessing and extracting webpage information, while text-image matching technology combines computer vision and natural language processing to enable computers to understand textual descriptions and retrieve relevant images. Notably, the CLIP model, a deep learning-based text-image matching technology, has garnered attention for its high accuracy, generalization ability, and unsupervised learning, which does not require large amounts of labeled data. The paper also introduces a major application in the realm of art and cultural heritage restoration, where these technologies play a pivotal role in the digital reconstruction, preservation, and virtual presentation of artworks and artifacts. Additionally, the importance of data ethics, privacy protection, algorithmic fairness, and preventing technological misuse are emphasized. The paper concludes with future trends pointing towards more intelligent web crawlers, more precise image semantic understanding, and more user-friendly experiences.