Typical Ceramic Pattern Extraction Based on Convolutional Neural Network

Pattern is the most fundamental dominant gene element in ceramics. In order to realize rapid batch extraction of ceramic patterns, establish the gene database of ceramic patterns, an automatic extraction method of ceramic patterns based on Convolutional Neural Network(CNN) was proposed. Taking Ming and Qing ceramics as an example, the experiment collected four typical patterns of Ming and Qing ceramics, including dragon patterns, fish patterns, peony patterns, and lotus patterns, and established four pattern datasets. Train based on four CNN models: PSPNet, DeepLabv3+, DANet, and HRNet. The experimental results show that HRNet and DANet have good extraction effects, which proves that Convolutional neural network is feasible for automatic extraction of pattern batch, and lays a foundation for further research.