Research on Classification of Clothing Fabrics Images Based on Convolutional Neural Network

Clothing fabrics are the material foundation of clothing design. In clothing design, different fabrics are usually used to compare visual effects. But it will take more time and increase material costs. And the choice of fabrics is also quite limited. Digitizing clothing fabrics and automatically categorizing them can solve these problems. The article established a dataset containing 15000 labeled high-definition clothing fabric images belonging to 10 fabric types. Then we use the VGG-16 network model that has been pre-trained on the ImageNet dataset to conduct transfer learning on the fabric images dataset. Finally, by analyzing the experimental results, we found that the VGG-16 model of transfer learning has a strong generalization ability, and its test accuracy in fabric images classification reached 99.73%. The trained convolutional neural network automatic classification model can be applied in digital clothing design, virtual fitting and other aspects. Finally, digital application of clothing fabrics will be achieved.