以類神經網路辨識服飾造型之研究

Applying an Artificial Neural Network to Recognition of Apparel Styling

林正中(J. J. Lin)

《紡織綜合研究期刊》 17卷1期 (2007/01) Pp. 47-58

https://doi.org/10.6439/TTRJ.200701.0047

摘要

本文應用類神經網路建構一智慧型服飾造型辨識系統，先以類神經網路學習不同的造型樣本，使系統具備服裝造型設計的專業知識及智慧以扮演專業設計人員提供消費者服飾造型搭配咨詢服務之角色。本研究首先廣泛收集有關服飾造型設計的樣本儲存於資料庫（造型感覺包括：摩登，端莊，輕便，運動四種；造型的搭配因子包括款式及配色），並將各種不同款式造型依其個別特性加以編碼轉換成特徵向量，以供類神經網路學習之用，使系統具備辨識服飾造型專業領域的知識。經由此造型辨識系統的協助，可提供消費者咨詢確認服飾造型感覺以供其參考。

關鍵字

類神經網路 ； 造型辨識 ； 搭配因子 ； 搭配屬性

並列摘要

The paper presents a styling classifying system developed by artificial neural network (ANN). It can play a role of styling consultant after learning various samples of apparel styling to be of expertise of styling. There are four basic selected apparel-styles, i.e., modern, elegant, leisure, and sporty, each of which is composed of components such as coat, shirt, vest, skirt, trousers and the colors adopted for them. The searched instances of different styling are deconstructed and encoded into a feature vector for training (Learning) of ANN. With the assistance of the system, consumers are able to classify the styling category for their dressing without any expertise of styling.

並列關鍵字

Artificial neural network ； styling recognition ； matching factors ； matching attributes