Guided Hash Algorithm for Information Semantic Retrieval in Multimedia Environment

Abstract

With technological progress, how to efficiently process and retrieve large amounts of multimodal multimedia data has become a challenge. Although multi-modal hash algorithms have been applied in this field, their performance potential has not been fully realized. To fully leverage the performance of the model, a supervised discrete multimodal hash algorithm has been proposed to improve the efficiency of multimodal retrieval. The adaptive online multi-modal hash algorithm is used to dynamically adapt to changes in query samples. The topological semantic multi-modal hash algorithm is applied to further improve the retrieval performance of multi-modal hashes. According to the results, in the MIR Flickr dataset, the Mean Average Precision (MAP) reaches 0.8048, 0.8155, and 0.8185 at 32-bit, 64bit, and 128-bit, respectively. Fusion Graph Convolutional Multimodal Hashing (FGCMH) exhibits the bestin different datasets. From this, the designed method has high processing power in handling large-scale and high-dimensional multimedia data.