Overview of Image Segmentation and Its Application on Free Space Detection

With the development of deep learning technique, image segmentation has received spreading attention in the computer vision field. It has a wide range of applications such as scene understanding, autonomous driving and so on. For the image segmentation, we can divide it into the semantic segmentation and instance segmentation, where a high-quality segmentation label for each instance is required for the latter method. In this paper, we sort out the popular structures of semantic segmentation and introduce the instance segmentation briefly. In the experiments, three main semantic segmentation methods are tested and analyzed based on the opened CamVid dataset, and the experiments for free space detection based on two popular segmentation methods are given.