請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/49721
標題: | 基於風格辨識之服裝推薦系統 Clothes Recommendation System based on Style Recognition |
作者: | Sz-Han Chao 趙思涵 |
指導教授: | 李明穗 |
關鍵字: | 服裝推薦,風格辨識,服裝參數,色票,構面情緒模型, clothes recommendation,style recognition,clothing attribute,color theme,dimension emotion model, |
出版年 : | 2016 |
學位: | 碩士 |
摘要: | 很多人每天都會花上許多時間選擇當天要穿著的服裝,對於他們來說,花許多時間選擇每天的穿搭是一向不小的困擾,為了幫助這些人更容易地選擇服裝,我們設計了一套依據風格的服裝推薦系統,可以推薦適當的穿搭。在我們的系統中,將衣服的風格定義為一種情緒,代表人們在看見衣服或是看見某個人穿著這件衣服之後的感覺,因此我們利用構面情緒模型(dimension emotion model)來設計我們的風格空間(style space)。除了使用低階特徵之外,我們使用服裝參數(clothing attribute)作為連接低階特徵與高階服裝風格的中階特徵,並利用支持向量回歸(support vector regression)來訓練風格分數函數(style score function)用於辨識服裝風格。我們使用風格色票(style color theme)來評估穿搭服裝的配色。因為沒有適合此系統的資料庫,我們建立了兩個新的資料庫,並使用我們的風格空間來標示其風格。經過實驗與使用者評估,以多數人的感官來說,此系統的推薦結果大部分都符合指定風格(情緒)。 Many people have trouble selecting clothes in the everyday morning. They often cost much time to choose the clothes. In order to help them easier choose clothes, we proposed a clothes recommendation system based on the style to recommend the suitable clothes for them. In our system, we define the style of clothes as the emotion. It means the feeling when people look at the clothes or look at a person wearing clothes. Thus, we use the dimension emotion model to design the style space. In addition to low-level features, the clothing attributes are used as the mid-level features. Then we train style score functions by the standard support vector regression to recognize the styles of clothes. Besides, the style color theme is used to assess the color matching of clothes. Because there is not suitable dataset for our system, we construct two new datasets with style labels by our style space. The experiments and user study show that the clothes recommended by our system almost match the style (emotion) in common sense. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/49721 |
DOI: | 10.6342/NTU201602497 |
全文授權: | 有償授權 |
顯示於系所單位: | 資訊工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-105-1.pdf 目前未授權公開取用 | 6.1 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。