Research on Digital Conservation and Restoration of Ancient Paintings Driven by Multi-modal Deep Learning

As an important carrier of cultural heritage, ancient Chinese paintings have gathered thousands of years of artistic wisdom and cultural essence, but are facing a serious inheritance crisis due to natural erosion and man-made destruction. With the popularity of digital technology, digital restoration technology is gradually replacing traditional manual restoration and becoming mainstream. This study focuses on the limitations of traditional digital technology in dealing with complex ancient paintings damage, multimodal data fusion, etc., and is committed to building an intelligent restoration system for complex damaged ancient paintings, and innovatively proposes to build a multimodal deep learning framework based on Transformer. The research breaks through the traditional single visual restoration mode, and realises the intelligent upgrading of ancient painting restoration through visual-semantic deep fusion.