ETRI-Knowledge Sharing Plaform

KOREAN
논문 검색
Type SCI
Year ~ Keyword

Detail

Conference Paper Deep Multi Class-wise Clothing Attributes Recognition for the Elderly Care Robot Environment
Cited - time in scopus Share share facebook twitter linkedin kakaostory
Authors
Chankyu Park, Minsu Jang, Jaeyeon Lee, Jaehong Kim
Issue Date
2020-09
Citation
International Symposium on Robot and Human Interactive Communication (RO-MAN) 2020, pp.1-4
Language
English
Type
Conference Paper
Abstract
In this study, we propose a multi-attribute-based deep neural network classification model that recognizes not only the basic clothing type but also the various sub-attributes of clothing in order to analyze clothing used as the most important factor when recognizing human appearance. This multi-attribute recognition model improves recognition performance by considering the association between multiple attribute values in order to improve classification performance with existing binary attributes. We use this technology for services where robots interact with the elderly in a home where robots care for the elderly. In particular, in order to reflect the characteristics of elderly people's dressing, images were collected from TV content that appeared a lot of elderly people and shops selling elderly people's clothes. Multiple attributes were defined with 13 top and bottom attributes and were used to train the multi class-wise model.
KSP Keywords
Attribute recognition model, Attribute-based, Binary attributes, Classification Performance, Classification models, Clothing attributes, Deep neural network(DNN), Elderly Care, Elderly People, Multiple attribute, Neural Network Classification