Search

In this paper, we focus on exploring effective methods for faster and accurate semantic segmentation. A common practice to improve the performance is to attain high-resolution feature maps with strong semantic representation. Two strategies are widely used: atrous convolutions and feature pyramid fusion, while both are either computationally intensive or ineffective. Inspired by the Optical Flow for motion alignment between adjacent video frames, we propose a Flow Alignment Module (FAM) to learn Semantic Flow between feature maps of adjacent levels and broadcast high-level features to high-resolution features effectively and efficiently. Furthermore, integrating our FAM to a standard feature pyramid structure exhibits superior performance over other real-time methods, even on lightweight backbone networks, such as ResNet-18 and DFNet. Then to further speed up the inference procedure, we also present a novel Gated Dual Flow Alignment Module to directly align high-resolution feature maps and low-resolution feature maps where we term the improved version network as SFNet-Lite. Extensive experiments are conducted on several challenging datasets, where results show the effectiveness of both SFNet and SFNet-Lite. In particular, when using Cityscapes test set, the SFNet-Lite series achieve 80.1 mIoU while running at 60 FPS using ResNet-18 backbone and 78.8 mIoU while running at 120 FPS using STDC backbone on RTX-3090. Moreover, we unify four challenging driving datasets (i.e., Cityscapes, Mapillary, IDD, and BDD) into one large dataset, which we named Unified Driving Segmentation (UDS) dataset. It contains diverse domain and style information. We benchmark several representative works on UDS. Both SFNet and SFNet-Lite still achieve the best speed and accuracy trade-off on UDS, which serves as a strong baseline in such a challenging setting. The code and models are publicly available at https://github.com/lxtGH/SFSegNets. [ABSTRACT FROM AUTHOR]

Chemical reaction kinetics can be evaluated by probing dynamic changes of chemical substrates or physical phenomena accompanied during the reaction process. Chemiluminescence, a light emitting exoenergetic process, involves random reaction positions and kinetics in solution that are typically characterized by ensemble measurements with nonnegligible average effects. Chemiluminescent reaction dynamics at the single-molecule level remains elusive. Here we report direct imaging of single-molecule chemiluminescent reactions in solution and probing of their reaction dynamics under catalytic conditions. Double-substrate Michaelis–Menten type of catalytic kinetics is found to govern the single-molecule reaction dynamics in solution, and a heterogeneity is found among different catalyst particles and different catalytic sites on a single particle. We further show that single-molecule chemiluminescence imaging can be used to evaluate the thermodynamics of the catalytic system, resolving activation energy at the single-particle level. Our work provides fundamental insights into chemiluminescent reactions and offers an efficient approach for evaluating catalysts.Research into the dynamics of chemical reactions at the single-molecule level is a pivotal undertaking. Here, the authors present a direct investigation of the chemiluminescent reaction dynamics of single molecules in solution, providing spatiotemporally resolved insights into chemical reactions. [ABSTRACT FROM AUTHOR]

This research proposed the modular prefabricated permanent formwork system made of ultra-high-performance concrete (UHPC). Two kinds of modular formwork shapes were designed: the flat formwork and the ribbed. The experimental investigation on the axial compression performance of the composite columns that consist of the normal strength concrete (NSC) core and the modular UHPC permanent formwork was demonstrated. Compared with the flat formwork, the ribbed formwork exhibited better bonding with the NSC core. As observed from the test results, the composite column with the ribbed formwork presented a similar axial behavior as the NSC column with a slight improvement in ultimate loads. Therefore, the modular UHPC ribbed permanent formwork could be regarded as the additional cover to the conventional NSC column. In addition, the finite element analysis (FEA) model was also developed to simulate the composite columns numerically. The predicted capacities agreed with the experimental results, which validated the numerical models. The crack pattern estimated by the FEA model revealed that the interaction between the permanent formwork and the inner concrete introduced many tiny cracks to the concrete core. However, as protected by the UHPC permanent formwork, the overall durability of the composite columns can still be enhanced. [ABSTRACT FROM AUTHOR]

In many scientific fields which rely on statistical inference, simulations are often used to map from theoretical models to experimental data, allowing scientists to test model predictions against experimental results. Experimental data is often reconstructed from indirect measurements causing the aggregate transformation from theoretical models to experimental data to be poorly-described analytically. Instead, numerical simulations are used at great computational cost. We introduce Optimal-Transport-based Unfolding and Simulation (OTUS), a fast simulator based on unsupervised machine-learning that is capable of predicting experimental data from theoretical models. Without the aid of current simulation information, OTUS trains a probabilistic autoencoder to transform directly between theoretical models and experimental data. Identifying the probabilistic autoencoder's latent space with the space of theoretical models causes the decoder network to become a fast, predictive simulator with the potential to replace current, computationally-costly simulators. Here, we provide proof-of-principle results on two particle physics examples, Z-boson and top-quark decays, but stress that OTUS can be widely applied to other fields. [ABSTRACT FROM AUTHOR]

To reduce the cost of lightweight concrete (LWC) partition panels and to address recycling concrete waste, this work utilized completely recycled fine aggregate (CRFA) to replace the natural fine aggregate and ceramsite in the preparation of LWC and LWC partition panels. To this end, an autoclave-free curing process and an air-entraining agent were used to prepare the CRFA-LWC. The workability, compressive strength, drying shrinkage, and pore structure of the CRFA-LWC and the performance of the CRFA-LWC partition panels were then investigated. The results show that the optimal ratio of the CRFA to the cement is 2.2 for the lightweight concrete, and the optimal panel cross section is a rounded rectangular one. All the pores in the CRFA-LWC have a diameter of smaller than 0.17 mm, and the diameter of 89% of them is less than 0.05 mm. In order to satisfy the drying shrinkage requirements stipulated by Chinese code JC/T 169-2016, the CRFA-LWC should be cured for at least 10 days. The economic analysis concludes that the material cost of CRFA-LWC is 40% lower than that of the autoclaved ceramsite concrete. In addition, utilizing CRFA in lightweight concrete can ease the shortage of natural aggregate. Highlights: CRFA is used for preparing the lightweight concrete partition panels. The optimal ratio of the CRFA to the cement is 2.2 for the lightweight concrete. CRFA lightweight concrete needs to be cured for at least 10 days. The cost of lightweight concrete is reduced by 40% by using CRFA. [ABSTRACT FROM AUTHOR]

We present a probabilistic deep learning methodology that enables the construction of predictive data-driven surrogates for stochastic systems. Leveraging recent advances in variational inference with implicit distributions, we put forth a statistical inference framework that enables the end-to-end training of surrogate models on paired input–output observations that may be stochastic in nature, originate from different information sources of variable fidelity, or be corrupted by complex noise processes. The resulting surrogates can accommodate high-dimensional inputs and outputs and are able to return predictions with quantified uncertainty. The effectiveness our approach is demonstrated through a series of canonical studies, including the regression of noisy data, multi-fidelity modeling of stochastic processes, and uncertainty propagation in high-dimensional dynamical systems. [ABSTRACT FROM AUTHOR]

Using an alternative source decomposition, we propose new exact boundary conditions on numerical boundary of a square lattice for out-of-plane motion over the whole space. A set of recurrence relations are found for the resulting kernel functions, hence allow their efficient and accurate evaluation with a system of ordinary differential equations. Stability of the boundary conditions is proved rigorously. Numerical results illustrate effective suppression for spurious wave reflection, and elimination of corner effects. This approach may be extended to other lattice structures and in higher dimensions. [ABSTRACT FROM AUTHOR]

Fractional derivatives play a very important role in modeling physical phenomena involving long-range correlation effects. However, they raise challenges of computational cost and memory storage requirements when solved using current well developed numerical methods. In this paper, the differential operator multiplication method is proposed to address the issues by considering a reaction-advection-diffusion equation with a fractional derivative in time. The linear fractional differential equation is transformed into an integer order differential equation by the proposed method, which can fundamentally fix the aforementioned issues for select fractional differential equations. In such a transform, special attention should be paid to the initial conditions for the resulting differential equation of higher integer order. Through numerical experiments, we verify the proposed method for both fractional ordinary differential equations and partial differential equations. [ABSTRACT FROM AUTHOR]

Motivated by biological evolution, this paper explains the rationality of Vision Transformer by analogy with the proven practical evolutionary algorithm (EA) and derives that both have consistent mathematical formulation. Then inspired by effective EA variants, we propose a novel pyramid EATFormer backbone that only contains the proposed EA-based transformer (EAT) block, which consists of three residual parts, i.e., Multi-scale region aggregation, global and local interaction, and feed-forward network modules, to model multi-scale, interactive, and individual information separately. Moreover, we design a task-related head docked with transformer backbone to complete final information fusion more flexibly and improve a modulated deformable MSA to dynamically model irregular locations. Massive quantitative and quantitative experiments on image classification, downstream tasks, and explanatory experiments demonstrate the effectiveness and superiority of our approach over state-of-the-art methods. E.g., our Mobile (1.8 M), Tiny (6.1 M), Small (24.3 M), and Base (49.0 M) models achieve 69.4, 78.4, 83.1, and 83.9 Top-1 only trained on ImageNet-1K with naive training recipe; EATFormer-Tiny/Small/Base armed Mask-R-CNN obtain 45.4/47.4/49.0 box AP and 41.4/42.9/44.2 mask AP on COCO detection, surpassing contemporary MPViT-T, Swin-T, and Swin-S by 0.6/1.4/0.5 box AP and 0.4/1.3/0.9 mask AP separately with less FLOPs; Our EATFormer-Small/Base achieve 47.3/49.3 mIoU on ADE20K by Upernet that exceeds Swin-T/S by 2.8/1.7. Code is available at https://github.com/zhangzjn/EATFormer. [ABSTRACT FROM AUTHOR]