Search

Many current works directly adopt multi-rate depth-wise dilated convolutions to capture multi-scale contextual information simultaneously from one input feature map, thus improving the feature extraction efficiency for real-time semantic segmentation. However, this design may lead to difficult access to multi-scale contextual information because of the unreasonable structure and hyperparameters. To lower the difficulty of drawing multi-scale contextual information, we propose a highly efficient multi-scale feature extraction method, which decomposes the original single-step method into two steps, Region Residualization-Semantic Residualization. In this method, the multi-rate depth-wise dilated convolutions take a simpler role in feature extraction: performing simple semantic-based morphological filtering with one desired receptive field in the second step based on each concise feature map of region form provided by the first step, to improve their efficiency. Moreover, the dilation rates and the capacity of dilated convolutions for each network stage are elaborated to fully utilize all the feature maps of region form that can be achieved.Accordingly, we design a novel Dilation-wise Residual (DWR) module and a Simple Inverted Residual (SIR) module for the high and low level network, respectively, and form a powerful DWR Segmentation (DWRSeg) network. Extensive experiments on the Cityscapes and CamVid datasets demonstrate the effectiveness of our method by achieving a state-of-the-art trade-off between accuracy and inference speed, in addition to being lighter weight. Without pretraining or resorting to any training trick, we achieve an mIoU of 72.7% on the Cityscapes test set at a speed of 319.5 FPS on one NVIDIA GeForce GTX 1080 Ti card, which exceeds the latest methods of a speed of 69.5 FPS and 0.8% mIoU. The code and trained models are publicly available.

Velocity gradient is the basis of many vortex recognition methods, such as Q criterion, $\Delta$ criterion, $\lambda_{2}$ criterion, $\lambda_{ci}$ criterion and $\Omega$ criterion, etc.. Except the $\lambda_{ci}$ criterion, all these criterions recognize vortices by designing various invariants, based on the Helmholtz decomposition that decomposes velocity gradient into strain rate and spin. In recent years, the intuition of 'no vortex in straight flows' has promoted people to analyze the vortex state directly from the velocity gradient, in which vortex can be distinguished from the situation that the velocity gradient has couple complex eigenvalues. A specious viewpoint to adopt the simple shear as an independent flow mode was emphasized by many authors, among them, Kolar proposed the triple decomposition of motion by extracting a so-called effective pure shearing motion; Li et al. introduced the so-called quaternion decomposition of velocity gradient and proposed the concept of eigen rotation; Liu et al. further mined the characteristic information of velocity gradient and put forward an effective algorithm of Liutex, and then developed the vortex recognition method. However, there is another explanation for the increasingly clear representation of velocity gradient, that is the local streamline pattern based on critical-point theory. In this paper, the tensorial expressions of the right/left real Schur forms of velocity gradient are clarified from the characteristic problem of velocity gradient. The relations between the involved parameters are derived and numerically verified. Comparing with the geometrical features of local streamline pattern, we confirm that the parameters in the right eigen-representation based on the right real Schur form of velocity gradient have good meanings to reveal the local streamline pattern. Some illustrative examples from the DNS data are presented., Comment: 16 pages, 9 figures, 1 appendix