Intelligent generation algorithm of ceramic decorative pattern
Xinxin Liu, Hua Huang, Hao Wu · 2020
Aim to resolve the problem of limited versatility of ceramic decorative pattern generation method. An intelligent generation algorithm of ceramic decorative patterns is proposed based on deep neural network learning, by using the data set about flowers prepared in advance to train the deeplab v3+ network model for semantic segmentation, along with the trained model to perform the image style transfer experiment, in order to generate a pattern similar to the decorative on ceramics. The experiment results show that the generated ceramic decorative pattern is closer to the real decorative pattern, and the effect is more realistic and practical. The proposed method enable to convert photos of ordinary flowers into ceramic decorative pattern, which has a better effect than using image style migration alone, and it also can realize the intelligent generation of ceramic decorative patterns on deep neural networks.