BriefGPT.xyz
Jul, 2020
转置器:利用特征图作为转置卷积滤波器的通用纹理合成技术
Transposer: Universal Texture Synthesis Using Feature Maps as Transposed Convolution Filter
HTML
PDF
Guilin Liu, Rohan Taori, Ting-Chun Wang, Zhiding Yu, Shiqiu Liu...
TL;DR
提出了一种基于转置卷积操作的新方法,通过对输入纹理的整个编码特征图作为转置卷积滤波器,以及捕捉自相关信息的特征相似性图作为转置卷积输入,从而在几乎实时的单次前向传递中一次性合成未见过的纹理。
Abstract
Conventional
cnns
for
texture synthesis
consist of a sequence of (de)-convolution and up/down-sampling layers, where each layer operates locally and lacks the ability to capture the long-term structural dependenc
→