Zhou Kun

TL;DR: In this article, a binocular animation stylized rendering method based on deep learning is proposed, which divides the rendering process into two layers: rendering an original image, a parallax image, an occlusion map and like scene information at the first layer; and performing after-processing optimization based on the CNN at the second layer.

TL;DR: In this paper, a real-time binocular stylization rendering method based on deep learning is proposed, where an image conversion network is trained for a fixed style at first, and when running is carried out, the network is decomposed into a coder part and a decoder part; when rendering is performed, information, such as an original image, a parallax image and a motion field, is rendered through a three-dimensional scene; and furthermore, stylization is performed through post-processing based on the image conversion networks, so that a complete binocular animation is