Ten-Million-Order Human Database for World-Wide Fashion Culture Analysis

摘要

The paper proposes a huge geo-tagged image database, referred to as the Fashion Culture DataBase (FCDB), which is considered to be a semi-automatic image collection in cloud-based services such as social network services (SNS). In the stage of database construction, we introduce the large-scale data collection, refinement, and representation for SNS-based analysis. The proposed FCDB consists of 25,707,690 images for use in (i) inter-city similarity analysis and (ii) fashion trends visualization. We also present a simple but effective representation for the spatial and temporal analysis. Finally, we visualize an inter-city graph and the yearly trend change with the refined FCDB.