服装设计知识图谱中的服装装饰工艺分类模型

为解决服装装饰工艺类别判定中噪声视角或弱相关视角带来的负影响问题,采用具备视角约减功能的多视角分类模型对噪声视角或弱相关视角进行动态过滤。首先基于1-阶TSK模糊系统,在其目标优化函数中引入视角间分类误差一致性约束,实现多视角协同学习;然后通过变体信息熵学习各视角的权重,并在权重学习过程中制定视角约减规则,自动剔除噪声视角或弱相关视角;最后通过服装装饰工艺类别分类实验对所构建模型的分类精度进行验证。结果表明:相比视角约减之前,所提出的多视角分类模型的测试精度提高了2.68%,可有效地过滤噪声视角或弱相关视角,降低其对分类精度的影响。

Garment ornamenting craft classification model for knowledge graph on clothing design

In order to eliminate the negative influences caused by noisy views or weakly relevant views in the garment ornamenting craft classification tasks, an automatic view-reduction multi-view classification model was used to filter noisy views or weakly relevant views in this research. Based on the 1-order TSK fuzzy system, an error constraint item was introduced to be the objective function for collaborative learning. Then, a variant entropy item was designed to learn the weight of each view, and a reduction principle was designed to filter noisy views or weakly relevant views during collaborative learning. The proposed model was tested as the final step on the clothing ornamenting craft classification tasks. Experimental results demonstrate that the proposed classification model can reduce noisy views or weakly relevant views effectively such that the negative influences generated by them can be avoided. Compared with the model without view-reduction, the proposed classification model achieves a 2.68% improvement in terms of data accuracy.