2018年7月12日 · To this end, we present a semantically-consistent GAN framework, dubbed Sem-GAN, in which the semantics are defined by the class identities of image segments in the source domain as produced by a semantic segmentation algorithm.

In this paper, we propose a novel GAN architecture for pixel-level domain adaptation, coined semantically-consistent GAN (Sem-GAN), that takes as input two un-paired sets, each set consisting of tuples of images and their semantic segment labels, and learns a domain map-ping function by optimizing over the standard min-max generator-discriminato...

To this end, we present a semantically-consistent GAN framework, dubbed Sem-GAN, in which the semantics are defined by the class identities of image segments in the source domain as produced by a semantic segmentation algorithm.

2018年9月29日 · 我们设计了生成对抗网络（gan）来估计sem图像的噪声分布，然后使用生成器将学习的噪声添加到下采样的hr图像块并生成lr补丁。 我们在后面的讨论中将此模型称为Noise-GAN。

我们把这个扩展叫做半监督gan或sgan 论文实验结果表明，SGAN在 有限数据集 上比没有生成部分的基准分类器 提升了分类性能 。 论文实验结果表明，SGAN可以显著地 提升生成样本的质量 并 降低生成器的训练时间 。

To this end, we present a semantically-consistent GAN framework, dubbed Sem-GAN, in which the semantics are defined by the class identities of image segments in the source domain as produced by a semantic segmentation algorithm.

对于使用分类标签的’weakly-supervised’模型，GAN的生成器使用conditional GAN，将图片的分类标签作为输入。 作者在PASCAL VOC, SiftFLow, Stanford和CamVid等数据库上进行了实验。 其中在VOC上的实验结果如下： 可以看到在增加了大量无标注数据的情况下，作者提出的训练方法能够提升分割效果。 但是没有对比与’fully-supervised training + semi-supervised fine-tuning’框架的优劣。 因此实验结果只能体现文章创新点起到了作用，但是并没有体现有效的程度. CamVid: 可 …

Generative Adversarial Networks (GANs) have shown remarkable success in Semantic label map to Photo-realistic image Translation (S2PT) task. However, the result.

this end, we present a semantically-consistent GAN frame-work, dubbed Sem-GAN, in which the semantics are de-fined by the class identities of image segments in the source domain as produced by a semantic segmentation algorithm. Our proposed framework includes consistency constraints on the translation task that, together with the GAN loss

2017年10月11日 · A generative adversarial network (GAN) is designed to fit the noise of SEM images, and then generate realistic training samples from high resolution SEM data, and a fully convolutional network have been designed to perform image super-resolution and image denoise at the same time.