arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1901.06322cs.CVcs.LG
收藏

图像合成和图像到图像转换中的空间金字塔学习注意池

Learning Spatial Pyramid Attentive Pooling in Image Synthesis and Image-to-Image Translation

Department of ECE and the Visual Narrative Initiative, North Carolina State University(北卡罗来纳州立大学电气与计算机工程系及视觉叙事研究所)

Wei Sun, Tianfu Wu

摘要:

图像合成和图像到图像的翻译是两个重要的生成学习任务。通过分别学习生成对抗网络(GANs)~ \cite{goodfellow2014generative}和周期一致性GANs(CycleGANs)~ \cite{zhu2017unpaired},已经取得了显著的进展。本文提出了一种学习空间金字塔注意池(SPAP)的方法。SPAP是一种新的结构单元,可以很容易地集成到GANs和CycleGANs的生成器和鉴别器中。SPAP整合了Atrous空间金字塔\cite{chen2018deeplab}、级联注意力机制和剩余连接\cite{he2016deep}。它利用这三个组件的优点来促进有效的端到端生成学习:(i)通过ASPP融合多尺度信息的能力;(ii)通过注意力捕获空间位置(特别是多尺度上下文)或特征通道之间的相对重要性的能力;(iii)通过剩余连接保留信息并增强优化可行性的能力。粗到细和细到粗的SPAP进行了研究,并在这两个任务中观察到有趣的注意力地图。在实验中,所提出的SPAP在Celeba-HQ-128数据集~ \cite{karras2017progressive}上的GANs中进行了测试,并且在包括Cityscape数据集~ \cite{cordts2016cityscapes}、Facade和Aerial Maps数据集~ \cite{zhu2017unpaired}的图像到图像翻译数据集上的CycleGANs中进行了测试,两者都获得了更好的性能。

英文摘要:

Image synthesis and image-to-image translation are two important generative learning tasks. Remarkable progress has been made by learning Generative Adversarial Networks (GANs)~\cite{goodfellow2014generative} and cycle-consistent GANs (CycleGANs)~\cite{zhu2017unpaired} respectively. This paper presents a method of learning Spatial Pyramid Attentive Pooling (SPAP) which is a novel architectural unit and can be easily integrated into both generators and discriminators in GANs and CycleGANs. The proposed SPAP integrates Atrous spatial pyramid~\cite{chen2018deeplab}, a proposed cascade attention mechanism and residual connections~\cite{he2016deep}. It leverages the advantages of the three components to facilitate effective end-to-end generative learning: (i) the capability of fusing multi-scale information by ASPP; (ii) the capability of capturing relative importance between both spatial locations (especially multi-scale context) or feature channels by attention; (iii) the capability of preserving information and enhancing optimization feasibility by residual connections. Coarse-to-fine and fine-to-coarse SPAP are studied and intriguing attention maps are observed in both tasks. In experiments, the proposed SPAP is tested in GANs on the Celeba-HQ-128 dataset~\cite{karras2017progressive}, and tested in CycleGANs on the Image-to-Image translation datasets including the Cityscape dataset~\cite{cordts2016cityscapes}, Facade and Aerial Maps dataset~\cite{zhu2017unpaired}, both obtaining better performance.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑