深度学习下椅子造型特征标签识别与智能设计  

RECOGNITION AND INTELLIGENT DESIGN OF CHAIR SHAPE FEATURE LABEL UNDER DEEP LEARNING

作　　者：常健楠 李雪莲[1] CHANG JIANNAN;LI XUELIAN

机构地区：[1]浙江理工大学艺术与设计学院

出　　处：《设计》2024年第8期26-29,共4页Design

摘　　要：为了挖掘大数据图像的有效信息,准确识别海量照片中椅子造型特征并根据特征标签进行快速设计,研究一种基于深度学习的椅子造型特征智能设计方法。首先通过文献调研,确定基于类型学和形态分析法,为椅子造型特征要素选定标签;其次,运用MoblieNet网络构建大数据椅子图像的识别与分类模型,通过图像识别模型对获取的大批量椅子图像进行识别筛选,选择出符合要求的椅子图像;然后运用多标签学习模型建立椅子造型标签智能识别标注模型,验证模型有效性并识别标注椅子图像。最后,选定某标签的椅子图像可以构建训练数据集,通过生成对抗网络生成新图像,作为设计草图激活设计师灵感,更好地辅助智能设计。In order to mine the effective information of big data images,accurately identify chair modeling features in massive photos and quickly design them according to feature tags,an intelligent recognition method of chair modeling features based on deep learning is proposed.Firstly,through literature research,it is determined that the labels for the chair modeling feature elements are selected based on typology and morphological analysis.Secondly,MoblieNet network is used to build a recognition and classification model of big data chair images,and the large number of chair images obtained are identified and screened through the image recognition model,and the characters are selected.Meet the required chair image;Finally,the multi-label learning model is used to establish an intelligent recognition and labeling model for chair shapes,verify the effectiveness of the model,and identify the labeled chair images.Finally,the chair image with a certain label can be selected to build a training dataset,and a new image can be generated by generating an adversarial network,which can be used as a design sketch to activate designer inspiration and better assist intelligent design.

关 键 词：椅子设计 造型特征 深度学习 多标签学习 生成对抗网络 

分 类 号：TS655.4[轻工技术与工程]