Abstract
Recently, numerous handcrafted and searched networks have been applied forsemantic segmentation. However, previous works intend to handle inputs withvarious scales in pre-defined static architectures, such as FCN, U-Net, andDeepLab series. This paper studies a conceptually new method to alleviate thescale variance in semantic representation, named dynamic routing. The proposedframework generates data-dependent routes, adapting to the scale distributionof each image. To this end, a differentiable gating function, called softconditional gate, is proposed to select scale transform paths on the fly. Inaddition, the computational cost can be further reduced in an end-to-end mannerby giving budget constraints to the gating function. We further relax thenetwork level routing space to support multi-path propagations andskip-connections in each forward, bringing substantial network capacity. Todemonstrate the superiority of the dynamic property, we compare with severalstatic architectures, which can be modeled as special cases in the routingspace. Extensive experiments are conducted on Cityscapes and PASCAL VOC 2012 toillustrate the effectiveness of the dynamic framework. Code is available athttps://github.com/yanwei-li/DynamicRouting.