We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
少样本学习下的服装风格分析与评价.
- Authors
胡梦莹; 钟跃崎
- Abstract
In order to improve the objectivity of fashion style evaluation, a method of automatic feature extraction, recognition and classification of brand fashion image were proposed by using convolutional neural network. Different brands of fashion images were selected as the objects to explore the reasonable representation of their visual style and to classify the fashion style of different brands. A dataset established which contains 50 different kinds of brands fashion images, 30 for each category. 36 categories were randomly selected as training set, and the remaining 14 categories were used as support set and query set. Siamese network, Prototype network and Meta baseline network were used to test our data set. The classification results of each network model used were compared and analyzed. The results show that in the case of few images data, it can be classified by using few shot learning method. The classification accuracy of using meta baseline network in 5-way, 1-shot task is as high as 0. 947 5.
- Subjects
ARTIFICIAL neural networks; CONVOLUTIONAL neural networks; BRAND image; CLOTHING industry; BRAND identification; SIGNAL convolution
- Publication
Wool Textile Journal, 2021, Vol 49, Issue 4, p13
- ISSN
1003-1456
- Publication type
Article
- DOI
10.19333/j.mfkj.20210201205