Abstract
When we examine goods with their pictures at online shops, regarding flowers and drawings, it is difficult to give proper keywords for retrieving goods' pictures. To refer their pictures efficiently, image retrieval systems using adjectives called as Kansei words, which represent human feelings, impressions, etc., are proposed. However, since the current systems retrieve pictures without customers' preferences, outputs disagree with the preferences. To cope with the problem, we propose an image retrieval system with Kansei words according to users' preferences. Experimental results have shown the efficiencies of the proposed system comparing with performances of a conventional system.
