Image generation using generative adversarial networks and attention mechanism
Image generation using generative adversarial networks and attention mechanism
复制标题
DOI：
10.1109/icis.2016.7550880
复制
发表时间：
2016-06
期刊：
影响因子：
--
通讯作者：
Yuusuke Kataoka;Takashi Matsubara;K. Uehara
中科院分区：
文献类型：
--
作者：
Yuusuke Kataoka;Takashi Matsubara;K. Uehara
对于图像生成，训练深度神经网络以提取自然图像上的高级特征并从特征重建图像。然而，很难学会生成包含大量内容的图像。为了克服这个困难，已经提出了具有注意力机制的网络。它被训练来处理图像的部分并逐步生成图像。这使得网络能够处理图像的一部分的细节和整个图像的粗略结构。注意力机制由递归神经网络实现。此外，还提出了生成对抗网络（GANs）方法来生成更逼真的图像。在这项研究中，我们提出了利用注意力机制和GANs方法的有效性的图像生成。我们证明了我们的方法能够迭代构建图像，并且比标准GAN和DRAW的注意力机制更真实地生成图像。
For image generation, deep neural networks are trained to extract high-level features on natural images and to reconstruct the images from the features. However it is difficult to learn to generate images containing enormous contents. To overcome this difficulty, a network with an attention mechanism has been proposed. It is trained to attend to parts of the image and to generate images step by step. This enables the network to deal with the details of a part of the image and the rough structure of the entire image. The attention mechanism is implemented by recurrent neural networks. Additionally, the Generative Adversarial Networks (GANs) approach has been proposed to generate more realistic images. In this study, we present image generation where leverages effectiveness of attention mechanism and the GANs approach. We show our method enables the iterative construction of images and more realistic image generation than standard GANs and the attention mechanism of DRAW.