Considering the spectral properties of images, we propose a new self-attention mechanism with highly reduced computational complexity, up to a linear rate. To better preserve edges while promoting similarity within objects, we propose individualized processes over different frequency bands. In particular, we study a case where the process is merely over low-frequency components. By ablation study, we show that low frequency self-attention can achieve very close or better performance relative to full frequency even without retraining the network. Accordingly, we design and embed novel plug-and-play modules to the head of a CNN network that we refer to as FsaNet. The frequency self-attention 1) requires only a few low frequency coefficients as input, 2) can be mathematically equivalent to spatial domain self-attention with linear structures, 3) simplifies token mapping (1×1 convolution) stage and token mixing stage simultaneously. We show that frequency self-attention requires 87.29% ~ 90.04% less memory, 96.13% ~ 98.07% less FLOPs, and 97.56% ~ 98.18% in run time than the regular self-attention. Compared to other ResNet101-based self-attention networks, FsaNet achieves a new state-of-the-art result (83.0% mIoU) on Cityscape test dataset and competitive results on ADE20k and VOCaug. FsaNet can also enhance MASK R-CNN for instance segmentation on COCO. In addition, utilizing the proposed module, Segformer can be boosted on a series of models with different scales, and Segformer-B5 can be improved even without retraining. Code is accessible at https://github.com/zfy-csu/FsaNet.

In this article, based on a broad review of the literature, a comprehensive overview of the global state of the art in the development of rail vehicle crashworthiness is presented. Following large number of studies on rail vehicles, this article presents a review of the crashworthiness of passenger rail vehicles from the aspects of research approaches, accident investigations, mechanism analysis of occupant injuries and train damage, strategies of occupant protection, and crashworthiness design (crash energy management design) and performance of rail vehicles in collisions. Finally, some conclusions and recommendations for future research are presented.

Large-eddy simulations are used to investigate the origin of the wake asymmetry and symmetry behind notchback Ahmed bodies. Two different effective backlight angles, ${\beta _1} = 17.8\mathrm{^\circ }$ and ${\beta _2} = 21.0\mathrm{^\circ }$ , are simulated resulting in wake asymmetry and symmetry in flows without external perturbations, in agreement with previous experimental observations. In particular, the asymmetric case presents a bi-stable nature showing, in a random fashion, two stable mirrored states characterized by a left or right asymmetry for long periods. A random switch and several attempts to switch between the bi-stability are observed. The asymmetry of the flow is ascribed to the asymmetric separations and reattachments in the wake. The deflection of the near-wall flow structures behind the slant counteracting the asymmetry drives the wake to be temporarily symmetric, triggering the switching process of the bi-stable wake. The consequence of deflection that forces the flow structure to form on the opposite side of the slant is the decisive factor for a successful switch. Modal analysis applying proper orthogonal decomposition is used for the exploration of the wake dynamics of the bi-stable nature observed.