Automatic Ink Painting Rendering Technique Based on Deep Convolutional Neural Networks

Abstract

 Computers, due to their lack of a visual neural system, have struggled to enter the realm of art creation based on visual systems. However, this situation has been completely transformed with the emergence of deep convolutional neural networks. By constructing neural network models to extract style and texture features from artistic images, and then merging these features with content images, computers can generate content images with extraordinary artistic characteristics. The proposed neural network-based Deep Triangle Convolutional Network achieved a minimum testing time of 1.5 seconds and a classification error rate as low as 3.5%. This research was based on the artistic style algorithm of convolutional neural networks and developed and implemented a model for automatically rendering arbitrary images in ink painting style. The research results demonstrated that by implementing ink painting automatic rendering techniques based on deep convolutional neural networks, it can output images with distinctive ink painting features and digital image ink painting style rendering applications can be accomplished.

翻译

搜索

复制