基于Contextual Transformer的自动驾驶单目3D目标检测

针对当前单目3D目标检测中存在的漏检和多尺度目标检测效果不佳的问题,提出了一种基于Contextual Transformer的自动驾驶单目3D目标检测算法(CM-RTM3D)。在ResNet-50网络中引入Contextual Transformer(CoT),构建ResNet-Transformer架构以提取特征。设计多尺度空间感知模块(MSP),通过尺度空间响应操作改善浅层特征的丢失情况,嵌入沿水平和竖直两个空间方向的坐标注意力机制(CA),使用softmax函数生成各尺度的重要性软权重。在偏移损失中采用Huber损失函数代替L1损失函数。实验结果表明:在KITTI自动驾驶数据集上,相较于RTM3D算法,该算法在简单、中等、困难三个难度级别下,AP3D分别提升了4.84、3.82、5.36个百分点,APBEV分别提升了4.75、6.26、3.56个百分点。

Brain Functional Network Generation Using Distribution-Regularized Adversarial Graph Autoencoder with Transformer for Dementia Diagnosis

The topological connectivity information derived from the brain functional network can bring new insights for diagnosing and analyzing dementia disorders.The brain functional network is suitable to bridge the correlation between abnormal connectivities and dementia disorders.However,it is challenging to access considerable amounts of brain functional network data,which hinders the widespread application of data-driven models in dementia diagnosis.In this study,a novel distribution-regularized adversarial graph auto-Encoder(DAGAE)with transformer is proposed to generate new fake brain functional networks to augment the brain functional network dataset,improving the dementia diagnosis accuracy of data-driven models.Specifically,the label distribution is estimated to regularize the latent space learned by the graph encoder,which canmake the learning process stable and the learned representation robust.Also,the transformer generator is devised to map the node representations into node-to-node connections by exploring the long-term dependence of highly-correlated distant brain regions.The typical topological properties and discriminative features can be preserved entirely.Furthermore,the generated brain functional networks improve the prediction performance using different classifiers,which can be applied to analyze other cognitive diseases.Attempts on the Alzheimer’s Disease Neuroimaging Initiative(ADNI)dataset demonstrate that the proposed model can generate good brain functional networks.The classification results show adding generated data can achieve the best accuracy value of 85.33%,sensitivity value of 84.00%,specificity value of 86.67%.The proposed model also achieves superior performance compared with other related augmentedmodels.Overall,the proposedmodel effectively improves cognitive disease diagnosis by generating diverse brain functional networks.

The Similarity Transformation of Regularizing Functionals and Simularity Property of Their Fourier Transformations

Let and denote respectively the functionswhere λ≥1, The author discusses the similarity transformation of the regularizing functionals of these functions and the similar property of their Fourier transformation.

Depth-Guided Vision Transformer With Normalizing Flows for Monocular 3D Object Detection

Monocular 3D object detection is challenging due to the lack of accurate depth information.Some methods estimate the pixel-wise depth maps from off-the-shelf depth estimators and then use them as an additional input to augment the RGB images.Depth-based methods attempt to convert estimated depth maps to pseudo-LiDAR and then use LiDAR-based object detectors or focus on the perspective of image and depth fusion learning.However,they demonstrate limited performance and efficiency as a result of depth inaccuracy and complex fusion mode with convolutions.Different from these approaches,our proposed depth-guided vision transformer with a normalizing flows(NF-DVT)network uses normalizing flows to build priors in depth maps to achieve more accurate depth information.Then we develop a novel Swin-Transformer-based backbone with a fusion module to process RGB image patches and depth map patches with two separate branches and fuse them using cross-attention to exchange information with each other.Furthermore,with the help of pixel-wise relative depth values in depth maps,we develop new relative position embeddings in the cross-attention mechanism to capture more accurate sequence ordering of input tokens.Our method is the first Swin-Transformer-based backbone architecture for monocular 3D object detection.The experimental results on the KITTI and the challenging Waymo Open datasets show the effectiveness of our proposed method and superior performance over previous counterparts.

3D printing of poly(ethyleneimine)-functionalized Mg-Al mixed metal oxide monoliths for direct air capture of CO_(2)

Direct air capture(DAC)of CO_(2)plays an indispensable role in achieving carbon-neutral goals as one of the key negative emission technologies.Since large air flows are required to capture the ultradilute CO_(2)from the air,lab-synthesized adsorbents in powder form may cause unacceptable gas pressure drops and poor heat and mass transfer efficiencies.A structured adsorbent is essential for the implementation of gas-solid contactors for cost-and energy-efficient DAC systems.In this study,efficient adsorbent poly(ethyleneimine)(PEI)-functionalized Mg-Al-CO_(3)layered double hydroxide(LDH)-derived mixed metal oxides(MMOs)are three-dimensional(3D)printed into monoliths for the first time with more than 90%adsorbent loadings.The printing process has been optimized by initially printing the LDH powder into monoliths followed by calcination into MMO monoliths.This structure exhibits a 32.7%higher specific surface area and a 46.1%higher pore volume,as compared to the direct printing of the MMO powder into a monolith.After impregnation of PEI,the monolith demonstrates a large adsorption capacity(1.82 mmol/g)and fast kinetics(0.7 mmol/g/h)using a CO_(2)feed gas at 400 ppm at 25℃,one of the highest values among the shaped DAC adsorbents.Smearing of the amino-polymers during the post-printing process affects the diffusion of CO_(2),resulting in slower adsorption kinetics of pre-impregnation monoliths compared to post-impregnation monoliths.The optimal PEI/MeOH ratio for the post-impregnation solution prevents pores clogging that would affect both adsorption capacity and kinetics.

Changes in Transformation of Soil Organic C and Functional Diversity of Soil Microbial Community Under Different Land Uses

Changes in soil biological and biochemical properties under different land uses in the subtropical region of China were investigated in order to develop rational cultivation and fertilization management. A small watershed of subtropical region of China was selected for this study. Land uses covered paddy fields, vegetable farming, fruit trees, upland crops, bamboo stands, and forestry. Soil biological and biochemical properties included soil organic C and nutrient contents, mineralization of soil organic C, and soil microbial biomass and community functional diversity. Soil organic C and total N contents, microbial biomass C and N, and respiration intensity under different land uses were changed in the following order： paddy fields （and vegetable farming） 〉 bamboo stands 〉 fruit trccs （and upland）. The top surface （0-15 cm） paddy fields （and vegetable farming） were 76.4 and 80.8% higher in soil organic C and total N contents than fruit trees （and upland） soils, respectively. Subsurface paddy soils （15-30 cm） were 59.8 and 67.3% higher in organic C and total N than upland soils, respectively. Soil microbial C, N and respiration intensity in paddy soils （0-15 cm） were 6.36, 3.63 and 3.20 times those in fruit tree （and upland） soils respectively. Soil microbial metabolic quotient was in the order： fruit trees （and upland） 〉 forestry 〉 paddy fields. Metabolic quotient in paddy soils was only 47.7% of that in fruit tree （and upland） soils. Rates of soil organic C mineralization during incubation changed in the order： paddy fields 〉 bamboo stands 〉 fruit trees （and upland） and soil bacteria population： paddy fields 〉 fruit trees （and upland） 〉 forestry. No significant difference was found for fungi and actinomycetes populations. BIOLOG analysis indicated a changing order of paddy fields 〉 fruit trees （and upland） 〉 forestry in values of the average well cell development （AWCD） and functional diversity indexes of microbial community. Results also showed that the conversion from paddy fields to vegetable farming for 5 years resulted in a dramatic increase in soil available phosphorus content while insignificant changes in soil organic C and total N content due to a large inputs of phosphate fertilizers. This conversion caused 53, 41.5, and 41.3% decreases in soil microbial biomass C, N, and respiration intensity, respectively, while 23.6% increase in metabolic quotient and a decrease in soil organic C mineralization rate. Moreover, soil bacteria and actinomycetes populations were increased slightly, while fungi population increased dramatically. Functional diversity indexes of soil microbial community decreased significantly. It was concluded that land uses in the subtropical region of China strongly affected soil biological and biochemical properties. Soil organic C and nutrient contents, mineralization of organic C and functional diversity of microbial community in paddy fields were higher than those in upland and forestry. Overuse of chemical fertilizers in paddy fields with high fertility might degrade soil biological properties and biochemical function, resulting in deterioration of soil biological quality.

An extended functional transformation method and its application in some evolution equations

In this paper, an extended functional transformation is given to solve some nonlinear evolution equations. This function, in fact, is a solution of the famous KdV equation, so this transformation gives a transformation between KdV equation and other soliton equations. Then many new exact solutions can be given by virtue of the solutions of KdV equation.

Mechanisms of Transforming CH_x to CO on Ni(111) Surface by Density Functional Theory

To elucidate feasible routes of producing CO from CH 3 and unravel the eff ect of adsorbed O on CH x transformation, the reactivity of CH x ( x = 1-3) with and without the assistance of adsorbed atomic O on Ni(111) was explored using density functional theory calculations. The adsorption energies of CH x ( x = 0-3) were found to be signifi cantly reduced on an O-preadsorbed Ni(111) surface compared to a pure surface. Furthermore, O-assisted one-step dehydrogenation of CH x ( x = 1-3) features energy barriers and thus is diffi cult to proceed. In terms of energy, the direct dissociation of CH 3 is favorable, except for the last CH dehydrogenation, which is energy intensive. Interestingly, in O-assisted two-step CH transformation to CO via CHO intermediate, the barrier is dramatically lowered. The successive dehydrogenations of CH x O ( x = 1-3) were also found to be a route for CO formation. Finally, two possible pathways from CH 3 to CO are proposed:(a) CH 3 → CH 2 → CH → CHO → CO;(b) CH 3 → CH 3 O → CH 2 O → CHO → CO.

The Functional Gradient Description Method of Space Coordinate Transformation

The 3-dimension coordinate datum transformation is very important in geomatics.Generally,there are small angle coordinate transformation model (i.e.,Bursa model,et al) and large angle coordinate transformation model (i.e.,Rodrigo matrix model,et al).However,these papers only focus on the fixed and static situation regardless of the small or large angle.This paper proposes the conception of coordinate transformation gradient field,which can realize the space coordinate transformation from small angle to arbitrary angle,and from static to dynamical.Based on the equivalent characteristics of the unit quaternion rotation matrix and the Rodrigo matrix,through studying the mathematical relationship between the spatial coordinate transformation and the functional gradient,we derive a functional gradient expression of the arbitrary transformation formula in space.The study shows that the essence of spatial coordinate transformation is a kind of potential field mathematically.This potential field conception can unify all the space coordinate transformations.It is the theoretical foundation for the further study of time continuous space coordinate transformation,and this study gives a new solution for the attitude determination of motion carriers.

Transfer Function Based Recognition of Resonance Overvoltages Transformer Windings

Power transformer insulation systems are subjected to many stresses during normal operation due to lightning and switching.If the spectrum of incoming surge voltage matches the winding one,the corresponding resonance will be excited.Therefore external transients occurring in power systems might trigger internal overvoltages with large maximum value in transformer windings.Overvoltages having such characteristic have been the root cause of many power transformer failures.The paper presents an approach to the identification of sensitive zones in the transformer windings based on the measurements of overvoltages inside windings and frequency dependences of admittance of the power transformer.The frequency characteristic of the transformer winding may determine those regions in the frequency spectrum.The presented approach might be used both for design optimization and diagnostics of distribution and power transformers.

The Dual Functionals for the Generalized Ball Basis of Wang-Said Type and Basis Transformation Formulas

The generalized Ball curves of Wang-Said type with a position parameter L not only unify the Wang-Ball curves and the Said-Ball curves, but also include several useful intermediate curves. This paper presents the dual functionals for the generalized Ball basis of Wang-Said type. The relevant basis transformation formulae are also worked out.

A callus transformation system for gene functional studies in soybean

Obtaining transgenic plants is a common method for analyzing gene function. Unfortunately, stable genetic transformation is difficult to achieve, especially for plants（e.g., soybean）, which are recalcitrant to genetic transformation. Transient expression systems, such as Arabidopsis protoplast, Nicotiana leaves, and onion bulb leaves are widely used for gene functional studies. A simple method for obtaining transgenic soybean callus tissues was reported recently. We extend this system with simplified culture conditions to gene functional studies, including promoter analysis, expression and subcellular localization of the target protein, and protein-protein interaction. We also evaluate the plasticity of this system with soybean varieties, different vector constructs, and various Agrobacterium strains. The results indicated that the callus transformation system is efficient and adaptable for gene functional investigation in soybean genotype-, vector-, and Agrobacterium strain-independent modes. We demonstrated an easy set-up and practical homologous strategy for soybean gene functional studies.

Transformation of ammonium nitrogen and response characteristics of nitrifying functional genes in tannery sludge contaminated soil

High concentrations of ammonium nitrogen released from tannery sludge during storage in open air may cause nitrogen pollution to soil and groundwater.To study the transformation mechanism of NH_(4)^(+)-N by nitrifying functional bacteria in tannery sludge contaminated soils,a series of contaminated soil culture experiments were conducted in this study.The contents of ammonium nitrogen(as NH_(4)^(+)-N),nitrite nitrogen(as NO_(2)^(−)-N)and nitrate nitrogen(as NO_(3)^(−)-N)were analyzed during the culture period under different conditions of pollution load,soil particle and redox environment.Sigmodial equation was used to interpret the change of NO_(3)^(−)-N with time in contaminated soils.The abundance variations of nitrifying functional genes(amoA and nxrA)were also detected using the real-time quantitative fluorescence PCR method.The results show that the nitrification of NH_(4)^(+)-N was aggravated in the contaminated silt soil and fine sand under the condition of lower pollution load,finer particle size and more oxidizing environment.The sigmodial equation well fitted the dynamic accumulation curve of the NO_(3)^(−)-N content in the tannery sludge contaminated soils.The Cr(III)content increased with increasing pollution load,which inhibited the reproduction and activity of nitrifying bacteria in the soils,especially in coarse-grained soil.The accumulation of NO_(2)^(−)-N contents became more obvious with the increase of pollution load in the fine sand,and only 41.5%of the NH_(4)^(+)-N was transformed to NO_(3)^(−)-N.The redox environment was the main factor affecting nitrification process in the soil.Compared to the aerobic soil environment,the transformation of NH_(4)^(+)-N was significantly inhibited under anaerobic incubation condition,and the NO_(3)^(−)-N contents decreased by 37.2%,61.9%and 91.9%under low,medium and high pollution loads,respectively.Nitrification was stronger in the silt soil since its copy number of amoA and nxrA genes was two times larger than that of fine sand.Moreover,the copy numbers of amoA and nxrA genes in the silt soil under the aerobic environment were 2.7 times and 2.2 times larger than those in the anaerobic environment.The abundance changes of the amoA and nxrA functional genes have a positive correlation with the nitrification intensity in the tannery sludge-contaminated soil.

多尺度Transformer激光雷达点云3D物体检测

激光雷达点云3D物体检测,对于小物体如行人、自行车的检测精度较低,容易漏检误检,提出一种多尺度Transformer激光雷达点云3D物体检测方法 MSPT-RCNN(multi-scale point transformer-RCNN),提高点云3D物体检测精度。该方法包含两个阶段,即第一阶段(RPN)和第二阶段(RCNN)。RPN阶段通过多尺度Transformer网络提取点云特征,该网络包含多尺度邻域嵌入模块和跳跃连接偏移注意力模块,获取多尺度邻域几何信息和不同层次全局语义信息,生成高质量初始3D包围盒;在RCNN阶段,引入包围盒内的点云多尺度邻域几何信息,优化了包围盒位置、尺寸、朝向和置信度等信息。实验结果表明,该方法(MSPT-RCNN)具有较高检测精度,特别是对于远处和较小物体,提升更高。MSPT-RCNN通过有效学习点云数据中的多尺度几何信息,提取不同层次有效的语义信息,能够有效提升3D物体检测精度。

结合坐标Transformer的轻量级人体姿态估计算法

针对现有的大多数自底向上人体姿态估计算法存在模型规模大、计算成本高及对边缘设备不友好等问题,提出了一种基于YOLOv5s6-Pose的轻量级多人姿态估计网络模型YOLOv5s6-Pose-CT。该模型在颈部网络中引入空间和通道重建卷积,以减少空间和通道维度上的特征冗余。同时,提出了一种坐标Transformer嵌入于主干网络中,使模型专注于长距离依赖和拥有高效的局部特征提取能力。其次,通过使用无偏特征位置对齐来解决多尺度融合过程中出现的特征错位问题。最后,使用损失函数MPDIoU对边界框的回归损失重新定义。在COCO 2017数据集上的实验结果表明,本文优化的网络模型与主流的轻量级网络EfficientHRNet-H1模型相比,在保持相同精度的同时,参数量和计算量分别减少16.2%和66.1%。相比于基准模型YOLOv5s6-Pose,参数量减少11.2%,计算量降低5.8%,平均检测精度和平均召回率分别提升2.5%和2.6%。

基于Vision Transformer的时空卷积网络设计

目前主流人体动作识别大部分都是基于卷积神经网络(Convolutional Neural Network,CNN)实现,而CNN容易忽略视频中的空间位置信息,从而降低了视频空间频域中动作识别能力。同时传统CNN不能快速定位到关键的特征位置,并且在训练过程中不能并行计算导致效率低。为了解决传统CNN在处理时间频域和多并行计算问题,提出了基于视觉Transformer(Vision Transformer,ViT)和3D卷积网络学习时空特征(Learning Spatiotemporal Features with 3D Convolutional Network,C3D)的人体动作识别算法。使用C3D提取视频的多维特征图、ViT的特征切片窗口对多维特征进行全局特征分割;使用Transformer的编码-解码模块对视频中人体动作进行预测。实验结果表明,所提的人体动作识别算法在UCF-101、HMDB51数据集上提高了动作识别的准确率。

基于双向稀疏Transformer的多变量时序分类模型

针对多变量时序(Multivariate Time Series,MTS)分类中长序列数据难以捕捉时序特征的问题,提出一种基于双向稀疏Transformer的时序分类模型BST(Bidirectional Sparse Transformer),提高了MTS分类任务的准确度.BST模型使用Transformer框架,构建了一种基于活跃度得分的双向稀疏注意力机制.基于KL散度构建活跃度评价函数,并将评价函数的非对称问题转变为对称权重问题.据此,对原有查询矩阵、键值矩阵进行双向稀疏化,从而降低原Transformer模型中自注意力机制运算的时间复杂度.实验结果显示,BST模型在9个长序列数据集上取得最高平均排名,在临界差异图中领先第2名35.7%,对于具有强时序性的乙醇浓度数据集(Ethanol Concentration,EC),分类准确率提高30.9%.

A Strategic Study on Functional Transition of the University Park in China

After the summarization of the development of the university park in China,it is considered that there existed a large gap between the actuality and construction objective of the university park.In order to help the university park walk out of the development dilemma,it needs to conduct adjustment and transformation to realize sustainable and healthy development.Through thorough demonstration,suggestions and countermeasures for functional transition of the university park have been proposed from the perspective of functional localization,resources sharing,development mode and management system.In functional localization,education industry should be developed in new urban area;in resources sharing,the university park should be developed into an organic entity;in development mode,the university park should develop integratedly;in management system,it should achieve the transition from the traditional to innovative type.

结合Transformer与生成对抗网络的水下图像增强算法

由于水下环境的多样性和光在水中受到的散射及选择性吸收作用,采集到的水下图像通常会产生严重的质量退化问题,如颜色偏差、清晰度低和亮度低等,为解决以上问题,本文提出了一种基于Transformer和生成对抗网络的水下图像增强算法。以生成对抗网络为基础架构,结合编码解码结构、基于空间自注意力机制的全局特征建模Transformer模块和通道级多尺度特征融合Transformer模块构建了TGAN(generative adversarial network with transformer)网络增强模型,重点关注水下图像衰减更严重的颜色通道和空间区域,有效增强了图像细节并解决了颜色偏差问题。此外,设计了一种结合RGB和LAB颜色空间的多项损失函数,约束网络增强模型的对抗训练。实验结果表明,与CLAHE(contrast limited adaptive histogram equalization)、UDCP(underwater dark channel prior)、UWCNN(underwater based on convolutional neural network)、FUnIE-GAN(fast underwater image enhancement for improved visual perception)等典型水下图像增强算法相比,所提算法增强后的水下图像在清晰度、细节纹理和色彩表现等方面都有所提升,客观评价指标如峰值信噪比、结构相似性和水下图像质量度量的平均值分别提升了5.8%、1.8%和3.6%,有效地提升了水下图像的视觉感知效果。

The Fourier Transform of a Sort of Generalized Function

The Fourier transforms of functions X ln e are given. where λ is a complex constant (λ≠-1,-2. …), a is a complex number. m=0. 1. 2,…,and