Rendering method from three-dimensional model to two-dimensional image based on deep learning

Abstract

The invention discloses a rendering method from a three-dimensional model to a two-dimensional image based on deep learning. The rendering method comprises the steps: based on a contour rendering andstylizing method of deep learning, for the timeliness of existing contour extraction calculation and the complexity of calculation, constructing an appropriate deep learning network structure, and establishing mapping between a computer extraction line draft and a hand-drawn line draft for the data features of a style migration problem without contrast data; and based on a light and shadow rendering and stylizing method based on deep learning, based on an unsupervised training mode of a CycleGan network, constructing a style conversion network structure; carrying out unsupervised training on the network by using a large number of unpaired data sets, inputting the information such as hand-painted stylized line draft, materials, illumination and normal information of the three-dimensional model into the trained network, completing rendering, and generating a 2D picture or video with color filling and shadow expression more conforming to the hand-painted style. The rendering method can beapplied to animation production or game picture rendering.

Description

technical field [0001] The invention belongs to the two fields of computer vision and deep learning, and specifically relates to a rendering method from a three-dimensional model to a two-dimensional image based on deep learning. Background technique [0002] As an art form with painting and other plastic arts as the main means of expression, cartoon belongs to non-realistic rendering. In the production process of traditional cartoon animation, a lot of boring cartoon drawing work is concentrated on the assistant animator, so a natural idea is to use the computer to reduce the work of the assistant animator, so as to improve the production efficiency of cartoon animation. In the animation production system mainly based on two-dimensional cartoon animation design, shape interpolation and area automatic coloring are usually used to complete all or part of the assistant animator's work, so as to improve the efficiency of cartoon animation production. But this method still requ...

AI Technical Summary

Problems solved by technology

[0005] There is still a more serious problem in non-photorealistic rendering: the excessive lines are not as natural and charming as the artist's hand-painted drafts

However, the contour lines rendered by the computer are consistent in thickness and extremely accurate, but this also makes the picture lack the charm it should have.

Images