SwiftGraph

Existing storage systems apply simple metadata and hierarchical directory trees, which fail to meet the requirements of content-based queries (i.e., semantic queries) like finding semantically similar images with one specific image file. Therefore, we designed and implemented SwiftGraph, a system middleware that supported semantic metadata generated from deep learning hash. SwiftGraph applied an effective graph structure to organize the semantic hash code and provided fast content-based query interfaces. The performance evaluation on the different datasets demonstrated that the query precision and efficiency of SwiftGraph outran other searchable metadata paradigms.