Cultural Relic Image Retrieval Based on Artificial Intelligence

Cultural relic image retrieval is one of the key technologies in regulating the circulation of cultural relics, aiming to ensure the legitimacy of cultural relics being traded by analyzing the images and matching them with the database of cultural relics prohibited for sale or historical transactions. Traditional techniques present issues such as low processing efficiency, vulnerability to subjective influence, and inadequate retrieval performance. In this chapter, we introduce two deep cultural relic image retrieval algorithms to solve these problems. First, we propose a semantic a priori induced cultural relic image retrieval algorithm to deal with the large amount of subtle and complex discriminative information in the cultural relic images that are difficult to be effectively extracted and recognized by general deep image algorithms. Then, we introduce the sketch-based cross-domain cultural relic image retrieval task and propose the prototype contrastive learning algorithm for the cross-domain semantic gap challenges. Extensive experimental results show that our two deep algorithms have superior retrieval performance.