Fluid art control method based on deep learning

Abstract

The invention discloses a fluid art control method based on deep learning, and the method comprises the steps: a generation stage: enabling an original density field d of smoke to pass through an auto-encoder network with jump connection to generate a synthetic density field rendering stage, enabling the original density field d to pass through a micro renderer to generate a multi-angle original rendering image I [theta], a feature extraction stage in which the synthetic density field generates a multi-angle synthetic rendered image through a micro renderer, and an original rendered image I theta and a user-defined style image Is generate a reference rendered image Igt through a style migration network SANet to serve as a true value of the synthetic rendered image; respectively extracting features corresponding to the reference rendering image Igt, the synthetic rendering image and the style image Is through a pre-trained VGG network; finally, the extracted features participate in calculation of a loss function, the features of the synthetic rendering image are matched with the features of the reference rendering image and the features of the style image respectively by adjusting the features of the synthetic rendering image, and the fluid art generation effect is controlled.

Description

technical field [0001] The invention relates to the field of computer graphics and virtual simulation, and mainly relates to the field of fluid simulation. Specifically, the present invention provides a fluid art control method based on deep learning. Background technique [0002] Fluid art control is one of the important applications of fluid shape guidance and synthesis, and an important form of digital information production in the field of fluid simulation. For fluid simulation scenarios, the artistic control of the fluid not only preserves the physical laws of motion and flow characteristics, but also produces turbulent details and texture features in a specific artistic style. It performs artistic reprocessing of real-world fluids, changing the fluid structure and overall visual effects. The artistic control of fluids has been applied in aspects such as film special effects production, video game production, etc., which is of great significance to workers engaged in ...

AI Technical Summary

Problems solved by technology

[0007] Therefore, the following limitations still exist in the existing methods: 1) The artistic control of fluid still mainly relies on iterative optimization, and the operating efficiency is low; 2) The artistic editing of fluid mainly relies on very labor-intensive manual drawing, or pre-trained VGG network The method of spatially extracting style features, the gradient propagation process is more difficult

Images