Feature extraction for fashion trend detection

Author(s)
ABDELRAHMAN, SALMA
Advisor
Kim Dong Yoon
Department
일반대학원 컴퓨터공학과
Publisher
The Graduate School, Ajou University
Publication Year
2015-08
Language
eng
Keyword
computer visiondata miningimage miningimage recognitionfeature extraction
Abstract
Abstract This research work is aimed to digitally study and analyze fashion and cloths’ design by mining fashion images. The ultimate goal would be detecting fashion trends and find design and fashion patterns that will help in predicting fashion and the general public reaction and the level of success or failure certain designs and collections may face. The process to reach that goal involves studying existing algorithms and approaches for image processing and computer vision, as in texture analysis, image enhancement and color theory. And studying existing work on fashion analysis and fashion recommendation engines that serves E-commerce and online shopping sites, which developed algorithms for computer vision and feature extraction which specifically targets cloth and fashion images. And finally, work with the accumulative knowledge on fashion and cloths image analysis to develop and propose my own feature set that I think can reflect and represent deigns, and help transfer the visual content to an abstract data sets, that we can apply upon it different machine learning and data mining techniques to reach our original set of goals. So the main challenge in this research is to reach the best set of features that can be extracted from a huge set images, fashion runway images, which are take in different times, conditions and using different technologies with different backgrounds and resolutions.
URI
https://dspace.ajou.ac.kr/handle/2018.oak/13176
Fulltext

Appears in Collections:
Graduate School of Ajou University > Department of Computer Engineering > 3. Theses(Master)
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Browse