面向云服务平台的产品感性评价及标注模型

Evaluation and annotation model of product Kansei attributes on cloud service platform

为消除产品设计方案评价过程中个体评价标准的差异,同时进一步提升产品设计方案的评价效率和精度,提出一种基于YOLOv3和DFL-CNN的设计方案智能评价模型。该模型面向工业设计云服务平台,通过基于Darknet-53的YOLOv3算法实现对产品类别的自动标注。以吹风机产品为例,构建了感性语义评价数据集,并通过基于VGG-16的DFL-CNN算法对其造型语义进行细粒度分类,实现了产品类别及其造型感性语义的自动标注。该模型将设计评价问题转换为识别及分类问题,能够实现毫秒级的设计方案评价效率及95%以上的评价精度。最后,通过与不经过YOLOv3的DFL-CNN以及DFL-CNN的基础网络VGG-16进行对比实验,证明了所提模型的有效性及优越性。

To eliminate the differences of individual evaluation criteria in product Kansei attributes evaluation,and further improve the evaluation efficiency of product,an intelligent evaluation model based on YOLOv3 and DFL-CNN was proposed.The proposed model was oriented to cloud service platform of industrial design and divided into two steps:(1)YOLOv3 was used to locate the product and label its categories automatically;(2)the labeled images and its position coordinates were sent into DFL-CNN module to classify its Kansei attributes.A case study was provided to validate the proposed model.The proposed model transformed the design evaluation task into the recognition and classification task in the field of computer vision,and achieved over 95%accuracy in the binary and triple classification tasks.The elapsed time for this model on a GTX 1070 gpu was approximately 0.31 seconds.By comparing with other CNNs such as VGG-16,the validity and superiority of the proposed model were proved.