.png){kind=icon}
返回
Forest fog rendering using generative adversarial networks
.png){kind=icon}
PRE
AI
DOI：10.1007/s00371-021-02376-z.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Producing a physically realistic rendering of forests with fog involves the simulation of light diffusion in participating media. This volumetric phenomenon increases the realism of natural scenes in several fields, such as video games and flight simulators. However, this effect in a natural scene is characterized by complexity at all scales, representing a challenge for the computer graphics community. In this paper, we propose a novel approach based on a generative adversarial network to estimate fog in forest scenes. Our approach operates in two steps. The first step consists of training a generative adversarial network on a large dataset of synthetic images. Our network takes four images as input (normal map, depth map, albedo map and RGB map without fog), and it generates an estimation of the rendering forest with fog for output. A reference image conditions the input images to produce a better approximation. The second step consists of the production of realistic images with fog. Our technique can be generalized to high-frequency lighting effects (specularity and shadow), so it does not require any calculation in 3D space.
Keyword:
Forest rendering
Generative adversarial network
Natural scene lighting
Deep learning
期刊
IF:
2.9
论文数:
4.6K
被引数:
6.5K
机构
引用论文
分享收藏Stylistic scene enhancement GAN: mixed stylistic enhancement generation for 3D indoor scenes风格场景增强GAN: 3D室内场景的混合风格增强生成.png)
VISUAL COMPUTER
OAAI
VISUAL COMPUTERIF2.9
OAAI分享收藏S2RGAN: sonar-image super-resolution based on generative adversarial networkS2RGAN: 基于生成对抗网络的声纳图像超分辨率VISUAL COMPUTER
VISUAL COMPUTERIF2.9
PREAI
分享收藏
查看更多内容