基于SSD与FaceNet的人脸识别系统设计

人脸识别技术广泛应用于考勤管理、移动支付等智慧建设中。伴随着常态化的口罩干扰,传统人脸识别算法已无法满足实际应用需求,为此,本文利用深度学习模型SSD以及FaceNet模型对人脸识别系统展开设计。首先,为消除现有数据集中亚洲人脸占比小造成的类内间距变化差距不明显的问题,在CAS-IA Web Face公开数据集的基础上对亚洲人脸数据进行扩充;其次,为解决不同口罩样式对特征提取的干扰,使用SSD人脸检测模型与DLIB人脸关键点检测模型提取人脸关键点,并利用人脸关键点与口罩的空间位置关系,额外随机生成不同的口罩人脸,组成混合数据集;最后,在混合数据集上进行模型训练并将训练好的模型移植到人脸识别系统中,进行检测速度与识别精度验证。实验结果表明,系统的实时识别速度达20 fps以上,人脸识别模型准确率在构建的混合数据集中达到97.1%,在随机抽取的部分LFW数据集验证的准确率达99.7%,故而该系统可满足实际应用需求,在一定程度上提高人脸识别的鲁棒性与准确性。

基于YOLO5Face重分布的小尺度人脸检测方法

针对复杂场景下小尺度人脸检测精度较低的问题,提出了一种基于YOLO5Face重分布的小尺度人脸检测方法。方法以YOLO5Face为基础,在网络浅层引入改进的CBAM注意力并对模型计算重分布,提升复杂场景下小尺度人脸检测精度的同时降低模型参数量;采用融合mixup的数据增强方法,充分训练模型小尺度人脸检测分支;依据人脸检测特性,将softmax损失作为分类损失以最大化类间特征的差异。在WiderFace各个子集上的实验结果表明,与主流人脸检测方法相比,改进后的模型满足实时性的同时,小尺度人脸检测精度较高,其中Hard子集检测精度比YOLO5Face提升2个百分点。

Modification of Nano Tourmaline Surface Treatment Agent and Its Performance on Negative Ion Release

In this paper,a kind of wall fabric’s surface treatment agent modified with nonionic surfactant was reported.This surface treatment agent was prepared by using nano tourmaline powder dispersion in water with surfactant as dispersants by sand milling.Under the influence of different dispersants,the negative ions releasing amount of functional wall fabrics,the milling process and the storage stability of nano tourmaline powder dispersion were discussed.The results showed that nano tourmaline powder dispersion achieved the smallest average diameter of 44 nm and had best storage stability that the average diameter maintained below 200 nm in 17 days when the addition amount of dispersant was 20 percent of the tourmaline powders’weight.What is more,the quantity of negative ion releasing achieved 6500 ion/cm3 when addition amount of dispersant was 30 percent.This technique could be used to strengthen productivity of nano tourmaline powder dispersion.

VR/AR-AdaptFace:面向虚拟现实与增强现实的自适应多模态面部替换模型

随着VR/AR技术的迅猛发展,用户对于沉浸式体验的需求日益增长。同时,虚拟人脸技术亦趋成熟。基于此,本文探索将高度拟真的虚拟人脸融入VR/AR,以增强用户体验的自然度与沉浸感。然而,在虚拟数字人领域,图像生成及换脸技术在VR/AR环境下仍遇诸多挑战,尤其是唇形合成模型在动态场景及多语言环境下的性能需进一步优化。为解决上述问题,本文提出VR/AR-AdaptFace模型,一个面向虚拟现实与增强现实的自适应多模态面部替换方案。该模型由两大模块构成:“文颜绘真”模块,采用先进的文本至图像转换技术和特定类别先验保存策略,优化虚拟人脸生成,并通过注意力机制大幅提升图像质量;“语唇映生”模块,依托强大的生成器、唇形同步判别器及视觉质量判别器,实现语音与唇形的精准同步,为VR/AR场景中的动态交互带来更加逼真的体验。

Model prediction of the effect of in-mold electromagnetic stirring on negative segregation under bloom surface

Aiming at the problem of negative segregation under a bloom surface, a coupling macrosegregation model considering electromagnetic field, flow, heat, and solute transport was established based on the volume average method to study the effect of in-mold electromagnetic stirring(M-EMS) on the negative segregation under the bloom surface. In the model, the influence of dendrite structure on the flow and solute transport was described by the change of permeability. The model was validated by the magnetic induction intensity of M-EMS and carbon segregation experiment. The results show that the solute C in the solidified shell in the turbulent zone of the bloom undergoes two negative segregations, whereby the first is caused by nozzle jet, and the second by the M-EMS. The severities of the negative segregation caused by M-EMS at different currents and frequencies are also different, and the larger the current is, or the smaller the frequency is, the more serious will be the negative segregation.With the M-EMS, the solute C distribution in the liquid phase of the bloom is more uniform, but the mass fraction of C in the liquid phase is higher than that without M-EMS.

基于改进YOLOv5s-face的Face5系列人脸检测算法

针对人脸检测中小尺度人脸和遮挡人脸的漏检问题,提出了一种基于改进YOLOv5s-face(you only look once version 5 small-face)的Face5系列人脸检测算法Face5S(face5 small)和Face5M(face5 medium)。使用马赛克(mosaic)和图像混合(mixup)数据增强方法,提升算法在复杂场景下检测人脸的泛化性和稳定性;通过改进C3的网络结构和引入可变形卷积(DCNv2)降低算法的参数量,提高算法提取特征的灵活性;通过引入特征的内容感知重组上采样算子(CARAFE),提高多尺度人脸的检测性能;引入损失函数WIoUV3(wise intersection over union version 3),提升算法的小尺度人脸检测性能。实验结果表明,在WIDER FACE验证集上,相较于YOLOv5s-face算法,Face5S算法的平均mAP@0.5提升了1.03%;相较于先进的人脸检测算法ASFD-D3(automatic and scalable face detector-D3)和TinaFace,Face5M算法的平均mAP@0.5分别提升了1.07%和2.11%,提出的Face5系列算法能够有效提升算法对小尺度和部分遮挡人脸的检测性能,同时具有实时性。

Probabilistic analysis of tunnel face seismic stability in layered rock masses using Polynomial Chaos Kriging metamodel

Face stability is an essential issue in tunnel design and construction.Layered rock masses are typical and ubiquitous;uncertainties in rock properties always exist.In view of this,a comprehensive method,which combines the Upper bound Limit analysis of Tunnel face stability,the Polynomial Chaos Kriging,the Monte-Carlo Simulation and Analysis of Covariance method(ULT-PCK-MA),is proposed to investigate the seismic stability of tunnel faces.A two-dimensional analytical model of ULT is developed to evaluate the virtual support force based on the upper bound limit analysis.An efficient probabilistic analysis method PCK-MA based on the adaptive Polynomial Chaos Kriging metamodel is then implemented to investigate the parameter uncertainty effects.Ten input parameters,including geological strength indices,uniaxial compressive strengths and constants for three rock formations,and the horizontal seismic coefficients,are treated as random variables.The effects of these parameter uncertainties on the failure probability and sensitivity indices are discussed.In addition,the effects of weak layer position,the middle layer thickness and quality,the tunnel diameter,the parameters correlation,and the seismic loadings are investigated,respectively.The results show that the layer distributions significantly influence the tunnel face probabilistic stability,particularly when the weak rock is present in the bottom layer.The efficiency of the proposed ULT-PCK-MA is validated,which is expected to facilitate the engineering design and construction.

Electric Potential in Surface Produced Negative Ion Source with Magnetic Field Increasing Toward a Wall

Electric potential near a wall for plasma with the surface produced negative ions with magnetic field increasing toward a wall is investigated analytically. The potential profile is derived analytically by using a plasma-sheath equation, where negative ions produced on the plasma grid （PG） surface are considered in addition to positive ions and electrons. The potential profile depends on the amount and the temperature of the surface produced negative ions and the profile of the magnetic field. The negative potential peak is formed in the sheath region near the PG surface for the case of strong surface production of negative ions or low temperature negative ions. As the increase rate of the magnetic field near the wall becomes large, the negative potential peak becomes small.

Ferromagnetic barrier-induced negative differential conductance on the surface of a topological insulator

The effect of the negative differential conductance of a ferromagnetic barrier on the surface of a topological insulat（ is theoretically investigated. Due to the changes of the shape and position of the Fermi surfaces in the ferromagnetic barrie the transport processes can be divided into three kinds： the total, partial, and blockade transmission mechanisms. The bias voltage can give rise to the transition of the transport processes from partial to blockade transmission mechanisms, which results in a considerable effect of negative differential conductance. With appropriate structural parameters, the currenl voltage characteristics show that the minimum value of the current can reach to zero in a wide range of the bias voltag and then a large peak-to-valley current ratio can be obtained.

Characteristic studies on positive and negative streamers of double-sided pulsed surface dielectric barrier discharge

The mechanisms of streamer generation and propagation in double-sided pulsed surface dielectric barrier discharge(SDBD)on both sides have been analyzed and investigated by experiment and numerical simulation.The fully exposed asymmetric SDBD has two discharge processes located on the high voltage electrode(HVE)side and the ground electrode(GE)side.Discharge images of the HVE side and GE side are taken by a digital camera under continuous pulse and ICCD(Intensified Charge Coupled Device)is utilized to diagnose the generation and propagation of streamers in single pulse discharge.In order to understand the physical mechanisms of streamer evolution more deeply,we establish a 2D simulation model and analyze it from the aspects of electron density,ion density,reduced electric field and electron impact ionization source term.The results show that the primary and secondary discharges on the HVE side and the GE side of the double-sided SDBD are composed of positive streamer and negative streamer,respectively.On the HVE side,the accumulation of positive charges on the dielectric surface causes the direction of the electric field to reverse,which is the principal factor for the polarity reversal of the streamer.On the GE side,both the negative charges accumulated on the dielectric surface and the falling voltage are the key factors for the streamer polarity switch.

Rock mass quality prediction on tunnel faces with incomplete multi-source dataset via tree-augmented naive Bayesian network

Rock mass quality serves as a vital index for predicting the stability and safety status of rock tunnel faces.In tunneling practice,the rock mass quality is often assessed via a combination of qualitative and quantitative parameters.However,due to the harsh on-site construction conditions,it is rather difficult to obtain some of the evaluation parameters which are essential for the rock mass quality prediction.In this study,a novel improved Swin Transformer is proposed to detect,segment,and quantify rock mass characteristic parameters such as water leakage,fractures,weak interlayers.The site experiment results demonstrate that the improved Swin Transformer achieves optimal segmentation results and achieving accuracies of 92%,81%,and 86%for water leakage,fractures,and weak interlayers,respectively.A multisource rock tunnel face characteristic(RTFC)dataset includes 11 parameters for predicting rock mass quality is established.Considering the limitations in predictive performance of incomplete evaluation parameters exist in this dataset,a novel tree-augmented naive Bayesian network(BN)is proposed to address the challenge of the incomplete dataset and achieved a prediction accuracy of 88%.In comparison with other commonly used Machine Learning models the proposed BN-based approach proved an improved performance on predicting the rock mass quality with the incomplete dataset.By utilizing the established BN,a further sensitivity analysis is conducted to quantitatively evaluate the importance of the various parameters,results indicate that the rock strength and fractures parameter exert the most significant influence on rock mass quality.

广角SS-OCT En Face结构投射图揭示未成年人玻璃体早期液化特征

目的:应用广角扫频源光学相干断层扫描成像(swept-source optical coherence tomography,SS-OCT)的en face结构投射图研究玻璃体早期液化特征。方法:使用SS-OCT进行18 mm×18 mm的容积(Cube)扫描,创建并分析健康未成年人(年龄5~18岁)70眼的系列玻璃体en face结构投射图。结果:在未成年人中,视网膜前的玻璃体包含4种液化结构,分别为后皮质前玻璃体囊袋(posterior precortical vitreous pocket,PPVP)、视盘前Martegiani区(the area of Martegiani,AM)、血管前液化裂隙(prevascular vitreous fissures,PVF)和液化池(cistern)。所有研究眼均能检出PPVP、AM和PVF,其中22眼(31.4%)的PPVP和AM连通。41眼(58.6%)可检出液化池,且其年龄大于未检出液化池的个体(P=0.01),液化池的发生与年龄呈正相关(r_(s)=0.315,P=0.008)。液化池的象限空间分布频率依次为颞上(90.2%)、鼻上(58.5%)、颞下(36.6%)、鼻下(24.4%),最常累及颞上象限(P<0.001)。结论:PPVP、AM和PVF是健康人群视网膜前玻璃体早期液化过程中均出现的特征。液化池的发生与年龄呈正相关,最常出现在颞上象限,可能是年龄相关性玻璃体液化变性的结果。

Effects of stress conditions on the generation of negative bias temperature instability-associated interface traps

The exponent n of the generation of an interface trap （Nit）, which contributes to the power-law negative bias temperature instability （NBTI） degradation, and the exponent’s time evolution are investigated by simulations with varying the stress voltage Vg and temperature T. It is found that the exponent n in the diffusion-limited phase of the degradation process is irrelevant to both Vg and T. The time evolution of the exponent n is affected by the stress conditions, which is reflected in the shift of the onset of the diffusion-limited phase. According to the diffusion profiles, the generation of the atomic hydrogen species, which is equal to the buildup of Nit, is strongly correlated with the stress conditions, whereas the diffusion of the hydrogen species shows Vg-unaffected but T-affected relations through the normalized results.

Narrowing the focus:Therapeutic cell surface targets for refractory triple-negative breast cancer

Triple-negative breast cancer(TNBC)is defined as a type of breast cancer with lack of expression of estrogen receptor,progesterone receptor and human epidermal growth factor 2 protein.In comparison to other types of breast cancer,TNBC characterizes for its aggressive behavior,more prone to early recurrence and a disease with poor response to molecular target therapy.Although TNBC is identified in only 25%-30%of American breast cancer cases annually,these tumors continue to be a therapeutic challenge for clinicians for several reasons:Tumor heterogeneity,limited and toxic systemic therapy options,and often resistance to current standard therapy,characterized by progressive disease on treatment,residual tumor after cytotoxic chemotherapy,and early recurrence after complete surgical excision.Cell-surface targeted therapies have been successful for breast cancer in general,however there are currently no approved cell-surface targeted therapies specifically indicated for TNBC.Recently,several cell-surface targets have been identified as candidates for treatment of TNBC and associated targeted therapies are in development.The purpose of this work is to review the current clinical challenges posed by TNBC,the therapeutic approaches currently in use,and provide an overview of developing cell surface targeting approaches to improve outcomes for treatment resistant TNBC.

Simulation research on surface growth process of positive and negative frequency detuning chromium atom lithographic gratings

Chromium atom photolithography gratings are a promising technology for the development of nanoscale length standard substances due to their high accuracy,uniformity,and consistency.However,the inherent difference between the interaction of positive and negative frequency detuning standing wave field and the atoms can cause a difference in the adjacent peak-to-valley heights of the grating in positive and negative frequency detuning chromium atom lithography,which greatly reduces its accuracy.In this study,we performed a controlled variable growth simulation using the semi-classical theoretical model and Monte Carlo method with trajectory tracking and ballistic deposition methods to investigate the influence of key experimental parameters on the surface growth process of positive and negative frequency detuning atomic lithography gratings.We established a theoretical model based on simulation results and summarized empirical equations to guide the selection of experimental parameters.Our simulations achieved uniform positive and negative frequency detuning atomic lithography gratings with a period of 1/4 of the wavelength corresponding to the atomic transition frequency,and adjacent peak-to-valley heights differing by no more than 2 nm,providing an important theoretical reference for the controllable fabrication of these gratings.

VanillaFaceNet:一种高精度快速推理的牛脸识别方法

快速精准确定牛只身份对于牛只活体贷款,改善牛只骗保等问题具有重要意义。针对不同牛只面部差异小,FaceNet网络层数深,推理速度较慢,模型分类精度不足等问题,该研究提出了基于FaceNet的牛脸识别方法-VanillaFaceNet。该方法首先将FaceNet的主干特征提取网络替换为极简网络VanillaNet-13并提出动态激活和增强型线性变换的激活函数两种方法提高网络的非线性;然后,提出一种新的DBCA(dual-branch coordinate attention)注意力模块,能够更好地反映不同牛只面部特征之间的差异,从而提高网络的识别精度;最后,针对triplet loss仅能减小牛只类间差异的问题,采用center-triplet loss联合监督来减少牛只类内差异,从而提高了相同牛只身份比对的准确性。基于自建的牛脸数据集对该模型进行训练和测试,试验结果表明,VanillaFaceNet对牛只识别的准确率达到88.21%,每秒传输帧数为26.23帧。与FaceNet、MobileFaceNet、CenterFace、CosFace和ArcFace算法相比,本文算法的识别准确率分别提高了2.99、9.58、6.26、3.85和4.49个百分点,推理速度分别提升了2.67、0.77、0.10、1.28和0.94帧/s。该模型对牛只有较为优秀的识别效果,适于在嵌入式设备上部署,实现了牛只面部识别精度和推理速度之间的平衡。

Numerical simulation of a truncated cladding negative curvature fiber sensor based on the surface plasmon resonance effect

A refractive index(RI)sensor based on the surface plasmon resonance effect is proposed using a truncated cladding negative curvature fiber(TC-NCF).The influences of the TC-NCF structure parameters on the sensing performances are investigated and compared with the traditional NCF.The simulation results show that the proposed TC-NCF RI sensor has an ultra-wide detection range from 1.16 to 1.43.The maximum wavelength sensitivity reaches 12400 nm/RIU,and the corresponding R^(2)of the polynomial fitting equation is 0.9999.The maximum and minimum resolutions are 2.56×10^(-5)and 8.06×10^(-6),respectively.In addition,the maximum amplitude sensitivity can reach-379.1 RIU^(-1)when the RI is chosen as 1.43.The proposed TC-NCF RI sensor could be useful in biochemical medicine,environmental monitoring,and food safety.

Climate Change Assessment Using Negative Extreme Deviations of Surface Temperature Field (Case Study: Tbilisi)

To conduct regional climate change assessment for the selected location (Tbilisi) more than century period hydrometeorological observation data of day-night minimal temperature is used. In the presented paper only temperature negative values of whole winter season are used. The temperature field change is characterized by the following four parameters: the minimal temperature sum (winter-day temperature is less than zero) of winter frost days;frost day number;season minimal temperature and the occurrence date of season minimal temperature. The modular structure of the above listed parameters has been studied and the mathematical expressions for temporal changes of those parameters have been obtained considering dynamical norms and cyclical changes. The intensity of this change in terms of global warming has been determined. One of the main parameters determining climate change is the variations of temperature field. Despite the fact that this parameter has been modified in the large range due to numerous impacts, they kept stable and provide sustainable equilibrium of the Earth’s energy potential during long-term (several decades) period. This was the reason why the Earth climate remained unchained. As it is known, the anthropogenic impact on the environment has breached the sustainable balance of the Earth energy potential, the potential has been gradually grown almost for more than a century and is known by the name of global warming. The above-mentioned process has already created the greatest threat to the existence of Earth’s biosphere and if it still continues the catastrophic results eventually have to be expected. This led that the problem solving has been set as an urgent task. The numerous fundamental studies for all regions of the world have been carried out to assess the intensity of this process.

Diversity and Distribution of Staphylococcal Chromosomal Cassettes Mec (SCCmec) Types I, II and III in Coagulase-Negative Staphylococcal Strains Isolated from Surfaces and Medico-Technical Materials of the University Hospital of Abomey-Calavi/Sô-Ava

The coagulase-negative staphylococci (CoNS) have long been considered to be low pathogenicity. The possibility of a horizontal transfer of resistance and virulence genes from S. aureus to CoNS could increase the pathogenicity of these bacteria. The objective of this work is to contribute to a better knowledge of the pathogenicity of (CoNS) strains isolated from surfaces and medico-technical materials of the University Hospital of Abomey-Calavi/Sô-Ava. Seventy strains of CoNS isolated from surfaces and medico-technical materials of the University Hospital of Abomey-Calavi were tested for methicillin resistance. The resistance to methicillin was evaluated phenotypically by the resistance of the strains to cefoxitin and then confirmed by the search for the mecA gene using PCR. The genes encoding staphylococcal chromosomal cassette (SCCmec) types I, II and III originally found in S. aureus were tested in CoNS by multiplex PCR using specific primers. All the strains studied showed resistance to methicillin. However, only 28.5% (20/70) carried the mecA gene. SCCmec was identified in only 17.14% (12/70) of these strains. Four strains carried mecA gene as well as one of the three types of SCCmec searched. SCCmec types I, II and III were identified in CoNS strains studied. SCCmec type I was the most frequent chromosomal cassette in mecA+ strains, only or in association with another SCCmec. The study also revealed methicillin-resistant strains carrying SCCmec lacking the mecA gene. Finally, 60% (12/20) of the strains were found to be non-typeable. Our results show that CoNS strains present a high resistance to methicillin and the source of this resistance in the CoNS of our study is not only the mecA gene. There is also a high diversity of SCCmec, justified by a large number of non-typeable CoNS strains. The mecA− SCCmec+ methicillin-resistant strains deserve to be sequenced for further studies.

A Privacy Preservation Method for Attributed Social Network Based on Negative Representation of Information

Due to the presence of a large amount of personal sensitive information in social networks,privacy preservation issues in social networks have attracted the attention of many scholars.Inspired by the self-nonself discrimination paradigmin the biological immune system,the negative representation of information indicates features such as simplicity and efficiency,which is very suitable for preserving social network privacy.Therefore,we suggest a method to preserve the topology privacy and node attribute privacy of attribute social networks,called AttNetNRI.Specifically,a negative survey-based method is developed to disturb the relationship between nodes in the social network so that the topology structure can be kept private.Moreover,a negative database-based method is proposed to hide node attributes,so that the privacy of node attributes can be preserved while supporting the similarity estimation between different node attributes,which is crucial to the analysis of social networks.To evaluate the performance of the AttNetNRI,empirical studies have been conducted on various attribute social networks and compared with several state-of-the-art methods tailored to preserve the privacy of social networks.The experimental results show the superiority of the developed method in preserving the privacy of attribute social networks and demonstrate the effectiveness of the topology disturbing and attribute hiding parts.The experimental results show the superiority of the developed methods in preserving the privacy of attribute social networks and demonstrate the effectiveness of the topological interference and attribute-hiding components.