基于GoogLeNet的色织物花型分类

Pattern Classification of Yarn-Dyed Fabrics Based on GoogLeNet

针对色织物花型人工视觉分类效率低的问题,提出了一种基于深度卷积神经网络(CNN)的色织物花型图像识别分类方法。构建了18种类别的色织物花型图像样本数据库,建立了基于GoogLeNet的色织物花型分类深度卷积神经网络,并通过实验分析选择最优的训练迭代期与学习率。结果表明利用深度神经网络分类识别色织物花型是可行、有效的。

Aiming at the low efficiency problem of manual classification for the pattern of yarn-dyed fabric, pattern recognition and classification method for yarn-dyed fabrics based on deep convolution neural network （CNN） was proposed. 18 kinds of sample data- base of yarn-dyed fabric patterns were established. Deep convolution neural network of yarn-dyed fabric pattern classification models were developed based on GoogLeNet. Optimal training epoch periods and learning rates were selected through experimental analysis based on the model evaluation criterion. Results showed that it was feasible and effective to classify yarn-dyed fabric patterns by deep convolution neural networks.