基于PatchTracker的对抗补丁防御算法

基于深度神经网络的目标检测技术已经广泛应用于各领域,然而,通过对抗补丁攻击在图像中添加局部扰动,以此来误导深度神经网络,对基于目标检测技术的视觉系统构成了严重威胁。针对这一问题,利用对抗补丁和图像背景的语义差异性,提出了一种基于PatchTracker的对抗补丁防御算法,该算法由上游补丁检测器与下游数据增强模块组成。上游补丁检测器使用带有注意力机制的YOLOV5(you only look once-v5)确定对抗补丁所在位置,有助于提高对小尺度对抗补丁的检测精度;将检测区域用合适的像素值覆盖以抹除对抗补丁,上游补丁检测器不仅能够有效降低对抗样本的攻击性,而且不依赖大规模的训练数据;下游数据增强模块通过改进模型训练范式,提高下游目标检测器的鲁棒性;将抹除补丁后的图像输入经过数据增强的下游YOLOV5目标检测模型。在公开的TT100K交通标志数据集上进行了交叉验证,实验表明,与未采取防御措施相比,所提算法能够有效防御多种类型的通用对抗补丁攻击,在检测对抗补丁图像时的mAP(mean average precision)提高65%左右,有效地改善了小尺度对抗补丁的漏检情况。与现有算法比较,所提算法有效提高了神经网络在检测对抗样本时的准确率。此外,所提算法不涉及下游模型结构的修改,具有良好的兼容性。

Formation and ecological response of sand patches in the protection system of Shapotou section of the Baotou-Lanzhou railway,China

The development of bare patches typically signifies a process of ecosystem degradation.Within the protection system of Shapotou section of the Baotou-Lanzhou railway,the extensive emergence of bare sand patches poses a threat to both stability and sustainability.However,there is limited knowledge regarding the morphology,dynamic changes,and ecological responses associated with these sand patches.Therefore,we analyzed the formation and development process of sand patches within the protection system and its effects on herbaceous vegetation growth and soil nutrients through field observation,survey,and indoor analysis methods.The results showed that sand patch development can be divided into three stages,i.e.,formation,expansion,and stabilization,which correspond to the initial,actively developing,and semi-fixed sand patches,respectively.The average dimensions of all sand patch erosional areas were found to be 7.72 m in length,3.91 m in width,and 0.32 m in depth.The actively developing sand patches were the largest,and the initial sand patches were the smallest.Throughout the stage of formation and expansion,the herbaceous community composition changed,and the plant density decreased by more than 50.95%.Moreover,the coverage and height of herbaceous plants decreased in the erosional area and slightly increased in the depositional lobe;and the fine particles and nutrients of soils in the erosional area and depositional lobe showed a decreasing trend.In the stabilization phases of sand patches,the area from the inlet to the bottom of sand patches becomes initially covered with crusts.Vegetation and 0-2 cm surface soil condition improved in the erosional area,but this improvement was not yet evident in the depositional lobe.Factors such as disturbance,climate change,and surface resistance to erosion exert notable influences on the formation and dynamics of sand patches.The results can provide evidence for the future treatment of sand patches and the management of the protection system of Shapotou section of the Baotou-Lanzhou railway.

Scheme Based on Multi-Level Patch Attention and Lesion Localization for Diabetic Retinopathy Grading

Early screening of diabetes retinopathy(DR)plays an important role in preventing irreversible blindness.Existing research has failed to fully explore effective DR lesion information in fundus maps.Besides,traditional attention schemes have not considered the impact of lesion type differences on grading,resulting in unreasonable extraction of important lesion features.Therefore,this paper proposes a DR diagnosis scheme that integrates a multi-level patch attention generator(MPAG)and a lesion localization module(LLM).Firstly,MPAGis used to predict patches of different sizes and generate a weighted attention map based on the prediction score and the types of lesions contained in the patches,fully considering the impact of lesion type differences on grading,solving the problem that the attention maps of lesions cannot be further refined and then adapted to the final DR diagnosis task.Secondly,the LLM generates a global attention map based on localization.Finally,the weighted attention map and global attention map are weighted with the fundus map to fully explore effective DR lesion information and increase the attention of the classification network to lesion details.This paper demonstrates the effectiveness of the proposed method through extensive experiments on the public DDR dataset,obtaining an accuracy of 0.8064.

Unveiling the adaptation strategies of woody plants in remnant forest patches to spatiotemporal urban expansion through leaf trait networks

Background:With the expansion of urban areas,the remnants of forested areas play a crucial role in preserving biodiversity in urban environments.This study aimed to explore the impact of spatiotemporal urban expansion on the networks of leaf traits in woody plants within remnant forest patches,thereby enhancing our understanding of plant adaptive strategies and contributing to the conservation of urban biodiversity.Methods:Our study examined woody plants within 120 sample plots across 15 remnant forest patches in Guiyang,China.We constructed leaf trait networks (LTNs) based on 26 anatomical,structural,and compositional leaf traits and assessed the effects of the spatiotemporal dynamics of urban expansion on these LTNs.Results and conclusions:Our results indicate that shrubs within these patches have greater average path lengths and diameters than trees.With increasing urban expansion intensity,we observed a rise in the edge density of the LTN-shrubs.Additionally,modularity within the networks of shrubs decreased as road density and urban expansion intensity increased,and increases in the average path length and average clustering coefficient for shrubs were observed with a rise in the composite terrain complexity index.Notably,patches subjected to‘leapfrog’expansion exhibited greater average patch length and diameter than those experiencing edge growth.Stomatal traits were found to have high degree centrality within these networks,signifying their substantial contribution to multiple functions.In urban remnant forests,shrubs bolster their resilience to variable environmental pressures by augmenting the complexity of their leaf trait networks.

Treatment of 198 Cases of Thyroid Nodules with Sanying Capsule Combined with Xiaoying Patch

[Objectives]To observe the clinical efficacy of Sanying capsule combined with Xiaoying Patch in treating thyroid nodule(TN).[Methods]Two groups were treated similarly,with 200 cases in the control group undergoing basic treatment for 12 weeks and 198 cases in the observation group receiving Sanying capsule combined with Xiaoying patches for the same duration.The clinical symptoms,number of nodules,diameter of the largest nodule,and maximum reduction of nodules were observed before and after treatment.A control analysis was performed,and the underlying mechanisms were explored.[Results]The primary symptoms of the observation group exhibited a more favorable improvement than those of the control group.Additionally,the number of nodules decreased,the diameter of the largest nodule decreased,and the maximum reduction of nodules decreased in both groups following treatment.However,the observation group demonstrated a more pronounced improvement than the control group(P<0.05).[Conclusions]The combination of Sanying capsule and Xiaoying patch has been demonstrated to be an effective treatment for TN,with a high degree of reliability in terms of safety.

Deep Global Multiple-Scale and Local Patches Attention Dual-Branch Network for Pose-Invariant Facial Expression Recognition

Pose-invariant facial expression recognition(FER)is an active but challenging research topic in computer vision.Especially with the involvement of diverse observation angles,FER makes the training parameter models inconsistent from one view to another.This study develops a deep global multiple-scale and local patches attention(GMS-LPA)dual-branch network for pose-invariant FER to weaken the influence of pose variation and selfocclusion on recognition accuracy.In this research,the designed GMS-LPA network contains four main parts,i.e.,the feature extraction module,the global multiple-scale(GMS)module,the local patches attention(LPA)module,and the model-level fusion model.The feature extraction module is designed to extract and normalize texture information to the same size.The GMS model can extract deep global features with different receptive fields,releasing the sensitivity of deeper convolution layers to pose-variant and self-occlusion.The LPA module is built to force the network to focus on local salient features,which can lower the effect of pose variation and self-occlusion on recognition results.Subsequently,the extracted features are fused with a model-level strategy to improve recognition accuracy.Extensive experimentswere conducted on four public databases,and the recognition results demonstrated the feasibility and validity of the proposed methods.

Observing eddy dye patches induced by shear instabilities in the surf zone on a plane beach

The effects of surf zone eddy generated by alongshore currents on the deformation and transport of dye are still poorly understood,and related tracer release experiments are lacking.Therefore,a tracer release laboratory experiment was conducted under monochromatic,unidirectional incident waves with a large incident angle(30°)on a plane beach with a 1:100 slope in a large wave basin.A charge-coupled device suspended above the basin recorded the dye patch image.The evolution of eddy dye patch was observed and the transport and diffusion were analyzed based on the collected images.Subsequently,a linear instability numerical model was adopted to calculate the perturbation velocity field at the initial stage.The observation and image processing results show that surf zone eddy patches occurred and were separated from the original dye patches.Our numerical analysis results demonstrate that the structure of the perturbation velocity field is consistent with the experimental observations,and that the ejection of eddy patches shoreward or offshore may be ascribed to the double vortex.

Evaluation of the Efficacy of Auricular Acupoint Pressure Patch Combined with Modified Huangqi Decoction in Treating Diabetic Nephropathy

Objective:To explore the efficacy of auricular acupoint pressure patch combined with modified Huangqi Decoction in treating diabetic nephropathy.Methods:60 patients with diabetic nephropathy treated in our hospital from January 2021 to December 2022 were selected for this study.The patients were randomly divided into two groups using the random number table method,with 30 patients in each group.Among them,the control group was treated with conventional Western medicine,while the experimental group was treated with auricular acupoint pressure patches combined with modified Huangqi Decoction.The patients’fasting blood glucose(FPG),2-hour postprandial blood glucose(2hPG),glycated hemoglobin(HbA1c),urinary protein quantification,urea nitrogen(BUN),serum creatinine(SCr),and other indicators were detected and recorded before and after treatment.Results:Before treatment,there was no statistically significant difference in the FPG,the 2hPG,and the HbA1c between the two groups of patients(P>0.05);after treatment,the FPG,the 2hPG,and the HbA1c of the patients in the experimental group were significantly lower than those in the control group(P<0.05).Before treatment,there was no statistically significant difference in the urinary protein quantification,the BUN,and the SCr between the two groups of patients(P>0.05);after treatment,the urinary protein quantification,BUN,and SCr of the patients in the experimental group were significantly lower than those in the control group(P<0.05).The experimental group showed better improvement in symptoms such as fatigue,backache,and frequency of nocturia(P<0.05).Conclusion:Auricular acupoint pressure patch combined with modified Huangqi Decoction effectively treats diabetic nephropathy and it helps control blood sugar and renal function indicators and improve clinical symptoms,therefore improving the patients’quality of life.

一种基于图像patches分割的自监督生成对抗模型

近年来,通过将基于角度旋转的自监督模型引入到生成对抗网络中,一定程度上缓解了生成对抗模型稳定性不足与模式坍塌问题。但是,这种自监督模型基于全局信息对比,缺乏局部相关性,模型特征提取能力不足,导致生成图片的全局相似度较高而像素精度不足。因此,提出一种新型自监督生成对抗模型,将图像块(patches)引入到传统的角度旋转自监督模型中,通过对图像进行分块,将局部图像与整体图像作为输入,在旋转角度判别器的基础上新增位置判别器,在提高模型全局相似性的同时,增强了局部特征相关性。在CIFAR-10与STL-10数据集上进行测试,实验结果表明,基于图像patches分割的自监督生成模型比角度旋转自监督模型的效果更好。

A wearable sweat patch for non-invasive and wireless monitoring inflammatory status

Sweat diagnostics are being developed to provide insights into monitoring human health status using an accessi-bly non-invasive technique of sweat analysis[1-3].Abundant compositions,ranging from electrolytes and metabolites to large proteins,can be found in sweat,which have similar types of physiological biomarkers observed in the blood[1].Recent advances in flexible electronics[4-7]have transformed conventional laboratory tests into personalized sweat molecu-lar analysis that facilitates real-time sensing of target biomark-ers[3].

基于字体特征与多尺度PatchGAN的中文字体风格转换研究

针对现有中文字体风格转换方法生成字符图像质量低以及生成字体图像与目标字体图像风格不一致的问题,提出基于字体特征与多尺度patch生成对抗网络的中文字体风格转换方法.首先,根据字体特点设计两个特征提取网络,分别提取字体风格特征和字符内容特征;然后,将两个特征输入生成器,利用字体风格特征约束生成字符图像的风格,字符内容特征约束生成字符图像的字形;最后,将生成字符图像输入到多尺度patch判别器中,对生成结果的多尺度图像块判断真假.实验结果表明,所提方法有效提升了生成字符图像的质量以及与目标字体的风格一致性.

Effect of interactions of two holes on tensile behavior of patch repaired carbon/epoxy woven laminates

The conventional case of patch repair involves bonding a patch over single damage/hole in the laminate.This work investigates the effect of interaction of two holes on the tensile behavior patch repaired carbon epoxy woven laminates.The specimens of[0°/45°/45°/0°]laminates were repaired with adhesively bonded two-ply[45°]2external patches.Three different cases of drilled specimens were produced with different hole arrangements viz.specimens with single central hole(SH),with two holes aligned along the longitudinal axis(LH)and with two holes along transverse axis(TH).The two-hole specimens were repaired with two different types,i.e.single large patches(SP)and with the two smaller patches(DP)of combined bonding area equal to the single large patches.Digital image correlation(DIC)was employed to capture strain contours.The results reveal the difference in the load transfer through the patches depending upon the arrangement of holes.The TH repaired specimen exhibit significant load recovery(SP-32.75%,DP-34.62%)while the LH specimens result in very marginal(SP-6.11%,DP-4.10%)recovery compared to their drilled case.The TH specimen failed by crack growing through both the holes beneath the patch,while the LH specimens failed by the failure through only one hole.The use of single large patch over multiple holes and multiple small patches individually over each hole has no significant influence on load recovery.

Integrated of Proximity-Coupling High-Order Modes Patch Antenna with Solar Cells for Sustainable Communication

An integration of single-layer proximitycoupling patch antenna and solar cells with bandwidth enhancement and optical energy harvesting is proposed for sustainable communication.For this purpose,many dual-function components are selected for designing the miniaturized solar cell antenna.On the one hand,by greatly affecting the current flow of the rectangular patch,vias and proximity-coupling are introduced to control the resonance modes frequency and matching,respectively,for wideband application,and the radiation performance property can be achieved by high-order mode.On the other hand,vias and proximity-coupling are beneficial to complete direct-current(DC)loop of solar cell and improve compatibility of DC-RF(radio frequency),whereas a high-order mode is beneficial to increase the area of collected light energy.To prove the working principle,fabricated and manufactured solar cell antenna.The measured and simulated results illustrate that the solar cell antenna gain is raised to as high as 9.27 d Bi in4.37 to 5.06 GHz applied to fifth generation communication(5G).

Divergent responses of Picea crassifolia Kom.in different forest patches to climate change in the northeastern Tibetan Plateau

Global climate changes have significantly affected tree growth and forest structures and functions in some arid and semi-arid regions,which are becoming warmer and wetter.Due to natural factors such as climate and terrain,some tree species may form different forest patches at the edges of their distribution areas.However,how forest patches of various sizes respond to climate change is unclear.In this study,we collected 203 tree cores from six different sizes of forest patches at the edge of the distribution area of Picea crassifolia Kom.in the northeast Tibetan Plateau.And we used the dendrochronology method to study the response of tree growth and resilience in different forest patches to climate change from 1961 to 2020.We simultaneously measured the contents of nonstructural carbohydrates(NSC),total nitrogen and total phosphorus of tree needles.Our results showed that the growth of trees in small-and medium-size forest patches(0.8–18.6 ha)has increased significantly.The early growing season(May–July)minimum temperature was the most important climate factor driving the growth of small-and medium-sized patch trees.The early growing season maximum temperature was the most important climate factor that inhibited the growth of trees in the largest patches(362.8 ha).The growth of individual trees in medium forest patches was better and the correlation with annual minimum temperature,maximum temperature,precipitation,actual evapotranspiration,and palmer drought severity index was stronger.The higher NSC content,stronger photosynthesis,and higher nitrogen utilization efficiency in leaves might be one of the reasons for the better growth of trees in moderate forest patches.In extreme drought years,as the forest patch area increased,the overall trend of tree growth resistance showed a unimodal pattern,with the highest at a forest patch area of 7.1 ha,while the overall trend of tree growth recovery was opposite.Therefore,we should strengthen the management of trees in large forest patches to cope with climate change.

Dielectric Patch Resonator and Antenna

This paper presents an overview of dielectric patch(DP)antennas developed in recent years.The employed DP resonator composed of a DP and a bottom substrate is analyzed comprehensively here,enabling the easy realization of a quasi-planar DP antenna.It combines the dual advantages of the conventional microstrip patch(MP)antenna and dielectric resonator(DR)antenna in terms of profile,gain,bandwidth,radiation efficiency,and design freedom.Furthermore,the DP antenna inherits the multi-mode characteristic of the DR antenna,thus it has a large number of high-order modes,including TMmn mode and TEmn mode.The high-order modes are widely applied,for example,by combining with the dominant TM10 mode to expand the bandwidth,or selecting multiple higher-order modes to implement a high-gain antenna.Additionally,the non-radiation high-order modes are also utilized to produce natural radiation null in filtering antenna design.In this paper,the design theories and techniques of DP antenna are introduced and investigated,including calculation and control methods of the resonant mode frequencies,analysis of the radiation mechanism,and applications of the multi-mode characteristic.This overview could provide guidance for the subsequent antenna design,thus effectively avoid time-consuming optimization.

Is It Safe to Perform an Autologous Epidural Blood Patch on Patients with Underlying Spinal Stenosis or Lumbar Disc Disease? Case Report and Literature Review

The most common spinal pathology seen in the obstetric population is lumbar disc herniation. There is currently no literature documenting the safety of performing an epidural blood patch on obstetric patients with underlying spinal pathology. We present a case of a patient with known severe lumbar spinal stenosis with compressive radiculopathy who received a successful epidural blood patch without worsening her underlying neurologic symptoms. Epidural blood patches can be safely performed in this patient population. However, the anesthesiologist should be aware of the risk of potentially worsening preexisting neurological deficits. Thus, we advise caution prior to placing an epidural blood patch on these patients. The risks and benefits of the procedure should be carefully weighed and considered. It is important to have a thorough discussion with the patient regarding the risks of an epidural blood patch prior to performing the procedure.

SRResNet Performance Enhancement Using Patch Inputs and Partial Convolution-Based Padding

Due to highly underdetermined nature of Single Image Super-Resolution(SISR)problem,deep learning neural networks are required to be more deeper to solve the problem effectively.One of deep neural networks successful in the Super-Resolution(SR)problem is ResNet which can render the capability of deeper networks with the help of skip connections.However,zero padding(ZP)scheme in the network restricts benefits of skip connections in SRResNet and its performance as the ratio of the number of pure input data to that of zero padded data increases.In this paper.we consider the ResNet with Partial Convolution based Padding(PCP)instead of ZP to solve SR problem.Since training of deep neural networks using patch images is advantageous in many aspects such as the number of training image data and network complexities,patch image based SR performance is compared with single full image based one.The experimental results show that patch based SRResNet SR results are better than single full image based ones and the performance of deep SRResNet with PCP is better than the one with ZP.

New therapy for oral ulcer:Evodia rutaecarpa patch combined with acupoint application

Background:Oral ulcer is the most common and easily recurrent disease in stomatology,which influence the patients’communication,normal dietary,and sleep.Evodia rutaecarpa(ER)is a traditional Chinese herbal medicine recorded in ancient Chinese medical books,which has a medicinal history of more than 2,000 years.Clinically,the application of ER at plantar Yongquan point(KI1)is effective in the treatment of oral ulcer.The purpose of this study was to combine the modern transdermal drug delivery system with traditional Chinese medicine to develop the transdermal absorption patch of ER and apply it to the treatment of oral ulcer at Yongquan point of plantar.Methods:Firstly,the medicinal materials of ER were extracted and the extracted materials were prepared into dispersed ER patch.The formulation and preparation process were screened by orthogonal design method and single factor investigation method.The adhesive and transdermal properties of the patch were used as the evaluation index of the preparation.Secondly,Wistar rats were used as experimental animals to establish a rat model of mouth ulcers.Wistar rats were randomly divided into normal group,model group(A),model group(B),low dose group,medium dose group and high dose group.the efficacy of ER on rat’s oral ulcer model was evaluated through three aspect such as apparent index,pathological index and biochemical index.Results:The patch had suitable adhesion and good skin penetration,which was an effective treatment for oral ulcer.In vivo pharmacodynamic studies,compared with the normal group,the body mass and food intake of rats in each group after modeling decreased,the amount of drinking water increased,the tissue structure of oral mucosa was damaged,and the levels of inflammatory factors(TNF-α,IL-6)and malondialdehyde increased,the levels of anti-inflammatory factors(IL-10),cell growth factor(epidermal growth factor,TGF-β_(1))and superoxide dismutase decreased.Compared with the model group,the body weight and food intake of each dosing group increased,water consumption decreased,the oral mucosal tissue structure was more complete,the levels of TNF-α,IL-6 and malondialdehyde decreased,the levels of IL-10,epidermal growth factor,TGF-β1 and superoxide dismutase increased,and the changes of various indexes were dose-related.Conclusion:ER patch can inhibit inflammatory reaction,enhance the antioxidant defense ability of the body,and promote the repair of damaged oral mucosa,so as to play an effective role in the treatment of oral ulcer.

Optimized Neural Network-Based Micro Strip Patch Antenna Design for Radar Application

Microstrip antennas are low-profile antennas that are utilized in wireless communication systems.In recent years,communication engineers have been increasingly interested in it.Because of downsizing,novelty,and cost reduction,the number of wireless standards has expanded in recent years.Wideband tech-nologies have evolved in addition to analog and digital services.Radars necessi-tate antenna subsystems that are low-profile and lightweight.Microstrip antennas have these qualities and are suited for radars as an alternative to the bulky and heavyweight reflector/slotted waveguide array antennas.A perforated corner single-line fed microstrip antenna is designed here.When compared to the basic square microstrip antenna,this antenna has better specifications.Because key issue is determining the best values for various antenna parameters when devel-oping the patch antenna.Optimized Neural Network(ONN)is one potential tech-nique utilized to solve this issue,and this work also uses Particle Swarm Optimization(PSO)to enhance the antenna performance.Return loss(S11)and Voltage Standing Wave Ratio(VSWR)parameters are considered in all situations,developed with Advanced Design System(ADS)applications.The transmitters are made to emit in the Ku-band,which covers a wide range of wavelengths.From 5–15 GHz,it is used in most current radars.The ADS suite is used to create the simulation design.

Epidural Blood Patches Performed with Miethke Sensor Reservoir for Continuous Intracranial Pressure Monitoring

An epidural blood patch (EBP) is a procedure performed by injecting autologous blood into a patient’s epidural space, usually at the site of a suspected CSF leak. It is typically performed in patients with characteristic postural headaches due to low intracranial pressure. We report a case of a young female with an implanted Miethke Sensor Reservoir, which was used for continuous intracranial pressure (ICP) monitoring during a two-level epidural blood patch. ICP increased only with thoracic injection, suggesting thoracic EBP may have greater efficacy than lumbar EBP in treating SIH and PDPH when the site of CSF leak is unknown.