基于生成对抗网络的半监督图像语义分割

Semi-Supervised Semantic Segmentation Based on Generative Adversarial Networks

提出了一种基于生成对抗网络的语义分割模型,包括一个全卷积语义分割网络以及一个判别网络,其中语义分割网络负责生成与输入图像对应的语义分割图,判别网络负责检测分割图与真实标签的区别,以促使分割网络改进分割效果。为了更好的提取全局结构信息,语义分割网络中采用了金字塔池化模块,对不同规模的空间区域进行池化操作。另外,为了应对语义分割训练数据集人工标注成本过高的问题,利用判别网络生成伪标签协助语义分割网络进行训练,从而实现了半监督训练效果。模型在PASCAL VOC2012数据集中进行了测试,结果表明该模型在全监督和半监督条件下均优于已有方法。

In this paper,we use generative adversarial network(GAN)to improve semantic segmentation of images.The model is composed of a semantic segmentation network and a discriminant network,where the segmentation network responses for generating semantic segmentation result while the discriminant network responses for detecting the difference between the generated result and the labels on the global structure level and improving the segmentation effect.In order to extract context information,we adopt the spatial pyramid pooling module in the segmentation network,which could perform pooling operation on multiple levels of sub-regions.Meanwhile,in order to solve the problem of a large number of manual annotations needed in the semantic segmentation data set,we use the discriminant network to generate pseudo labels and realize semi-supervision in the training of the segmentation network.The model has been tested using PASCAL VOC2012 dataset,and the results show that supervised and semi-supervised approaches proposed in this paper are superior to the existing methods.