Self-Attention Generative Adversarial Networks
Self-Attention Generative Adversarial Networks
复制标题
DOI：
--
复制
发表时间：
2018-05
期刊：
影响因子：
--
通讯作者：
Han Zhang;I. Goodfellow;Dimitris N. Metaxas;Augustus Odena
中科院分区：
文献类型：
--
作者：
Han Zhang;I. Goodfellow;Dimitris N. Metaxas;Augustus Odena
在本文中，我们提出了自我注意生成性对抗网络(SAGAN)，它允许对图像生成任务进行注意力驱动的远程依赖建模。传统的卷积GAN仅根据低分辨率特征地图中的空间局部点来生成高分辨率细节。在Sagan中，可以使用来自所有要素位置的线索来生成细节。此外，鉴别器可以检查图像的远距离部分中的高细节特征是否彼此一致。此外，最近的工作表明，发电机调节会影响GaN的性能。利用这一见解，我们将光谱归一化应用于GaN生成器，并发现这改善了训练动力学。建议的Sagan实现了最先进的结果，在具有挑战性的ImageNet数据集上，将最佳发布的初始得分从36.8提高到52.52，并将Frechet初始距离从27.62降低到18.65。注意层的可视化显示，生成器利用与对象形状相对应的邻域，而不是固定形状的局部区域。
In this paper, we propose the Self-Attention Generative Adversarial Network (SAGAN) which allows attention-driven, long-range dependency modeling for image generation tasks. Traditional convolutional GANs generate high-resolution details as a function of only spatially local points in lower-resolution feature maps. In SAGAN, details can be generated using cues from all feature locations. Moreover, the discriminator can check that highly detailed features in distant portions of the image are consistent with each other. Furthermore, recent work has shown that generator conditioning affects GAN performance. Leveraging this insight, we apply spectral normalization to the GAN generator and find that this improves training dynamics. The proposed SAGAN achieves the state-of-the-art results, boosting the best published Inception score from 36.8 to 52.52 and reducing Frechet Inception distance from 27.62 to 18.65 on the challenging ImageNet dataset. Visualization of the attention layers shows that the generator leverages neighborhoods that correspond to object shapes rather than local regions of fixed shape.