Prediction of parameters in the laser fading process of denim using convolutional neural networks

Abstract

Laser technology has been widely used in ecological fading of denim. However, the setting of parameters in the laser fading process mainly depends on the subjective experience of workers. After many attempts, an approximate effect can hardly be obtained. There are various defects in this way such as low fading efficiency, poor fading effect and high production cost, etc. This study proposed a novel method for predicting the parameters of the denim laser fading process using convolutional neural network technology. It starts from data collection, and uses laser technology to select different combinations of laser parameters to conduct laser fading experiments on 100% cotton indigo denim fabrics, and obtains denim image datasets of different laser fading effects. The prediction model of parameters in the laser fading process of denim was established by learning from the dataset with the convolutional neural network. The model could intelligently generate the approximation parameter group according to the fading image and performed well on the validation dataset with finite prediction error. Based on the obtained parameters, it could be effectively used and dramatically reduced the manpower and material resources in setting up a denim laser fading process.