In the eld of computer vision, image semantic segmentation is an important research branch and it is also a challenging task. Applications such as autonomous driving, Unmanned Aerial Vehicle System (UAVS), and even virtual or augmented reality systems require accurate and ecient segmentation