Image semantic segmentation is an important branch in the field of AI. Traditional semantic segmentation algorithms are mostly specific to the problem, and there is no universal segmentation algorithm suitable for all images. But deep learning approaches can solve this problem. Recently convolutional neural network is a widely used model in image segmentation, target recognition and scene classification, and has achieved great success. This survey introduces what image semantic segmentation is, what the semantic segmentation approaches are, and the methods of image segmentation with CNN. Next, we present several data sets that are often used in image segmentation experiments. Finally, these deep learning algorithms of image segmentation are analyzed and compared appropriately.