A comparative study of CBIR descriptors on innovative application of fashion image

Major Fashion trends in the world have been set up from dominant groups of fashion players from the leading fashion countries which have soon become fashion trend setters affecting to fashion trend area in a wide range from important fashion books to major catwalk parades. However, the process of fashion trends setting in many countries still remain complex and unpredictable. In this research we have studied how fashion trends rose up in Thailand, the strong textile industry-based country, by using CBIR (Content-Based Image Retrieval) technique. The top five well-known descriptors were experimented with a set of fashion images in various brands, seasons, and styles from past five years in major catwalk. A set of retrieved images from each descriptor was then compared with the set of relevant images classified by fashion gurus in Thailand. Finally, these result in finding the best descriptor which is suitable to Thailand fashion industry and used for developing fashion trend tracing back application afterwards.