深度学习方法下椅子风格识别与智能设计

Chair Style Recognition and Intelligent Design Method Based on Deep Learning

为了准确识别家具风格并根据风格进行快速设计,提出一种基于深度学习的家具风格识别和智能设计方法。以椅子为例,通过专家评价和聚类分析归纳出4种家具风格,分别是东方风格、西方风格、现代风格、新科技风格;运用ResNet50网络构建椅子图像的智能识别分类模型,并对标准数据集进行训练和识别,得到95.7%的识别准确率;随机输入10个图像进行检测,结果验证该识别分类模型有效;运用图像识别分类模型对网络爬虫采集的13万个椅子图像进行风格识别,筛选出23 869个具有明显风格指向的图像数据;以东方风格为例,筛选出1 391个图像数据,采用生成对抗网络生成对应风格的椅子图像和视频,并基于生成图像进行设计,实现家具指定风格的自动生成设计。

In order to meet the requirements of precise recognition of furniture style and rapid design based on style in the field of furniture design,a new design method based on deep learning is proposed.Taking the chair as an example,four furniture styles are summarized and classified through expert group and cluster analysis,which are Oriental Style,Western Style,Modern Style and New Technology Style.Based on the image intelligent recognition and classification model constructed by ResNet50,training ex-periments are carried out on 1 873 optimal images. The experimental results show that the recognition ac-curacy is 95.7%. With 10 randomly inputting images,the detection results verify that the recognition and classification model is effective. The image recognition and classification model is used to recognize the style of 136 127 chair images collected by web crawlers,and 23 869 images with obvious style orientation are selected. Taking the Oriental style as an example,1 391 images are screened out,and the correspond-ing chair images and videos are generated by the generative confrontation network. The automatic genera-tion and design of the specified style of furniture is realize based on the generated images.