为解决由于无人机视角下毛竹林的形状和纹理复杂,现有方法在分割精度和鲁棒性方面表现不佳的问题,提出了一种应用跨领域适应和偏移量引导的毛竹林分割网络——BFSNet。以百山祖国家公园为试验区,利用无人机拍摄周边毛竹林图像构建数据...为解决由于无人机视角下毛竹林的形状和纹理复杂,现有方法在分割精度和鲁棒性方面表现不佳的问题,提出了一种应用跨领域适应和偏移量引导的毛竹林分割网络——BFSNet。以百山祖国家公园为试验区,利用无人机拍摄周边毛竹林图像构建数据集。为增强模型的特征提取能力,提出跨领域适应模块以有效利用源模型的强特征提取能力,并结合自主学习提取适用于毛竹林分割任务的特征,利用两者的优势进行互补。为提高模型对于不同形状毛竹林的识别和定位能力,结合可变形卷积的偏移量引导模块,引入可学习的偏移量参数,以适应不同形状的毛竹林目标。将BFSNet在DeepGlobe Land Cover Classification Challenge和自制数据集上进行模型训练和测试,并与多种主流图像分割方法进行对比。结果表明:BFSNet在交并比、Dice系数、精确率和召回率4项指标上均取得了最优的性能表现,分别获得了76.04%和71.93%的交并比。与多种主流的图像分割模型相比,BFSNet在毛竹林的分割效果方面表现最为出色,对毛竹林形状的精确建模能力能够有效地应对不同形态的毛竹林。展开更多
Synthetic speech detection is an essential task in the field of voice security,aimed at identifying deceptive voice attacks generated by text-to-speech(TTS)systems or voice conversion(VC)systems.In this paper,we propo...Synthetic speech detection is an essential task in the field of voice security,aimed at identifying deceptive voice attacks generated by text-to-speech(TTS)systems or voice conversion(VC)systems.In this paper,we propose a synthetic speech detection model called TFTransformer,which integrates both local and global features to enhance detection capabilities by effectively modeling local and global dependencies.Structurally,the model is divided into two main components:a front-end and a back-end.The front-end of the model uses a combination of SincLayer and two-dimensional(2D)convolution to extract high-level feature maps(HFM)containing local dependency of the input speech signals.The back-end uses time-frequency Transformer module to process these feature maps and further capture global dependency.Furthermore,we propose TFTransformer-SE,which incorporates a channel attention mechanism within the 2D convolutional blocks.This enhancement aims to more effectively capture local dependencies,thereby improving the model’s performance.The experiments were conducted on the ASVspoof 2021 LA dataset,and the results showed that the model achieved an equal error rate(EER)of 3.37%without data augmentation.Additionally,we evaluated the model using the ASVspoof 2019 LA dataset,achieving an EER of 0.84%,also without data augmentation.This demonstrates that combining local and global dependencies in the time-frequency domain can significantly improve detection accuracy.展开更多
GEMIN5 is a predominantly cytoplasmic multifunctional protein, known to be involved in recognizing snRNAs through its WD40 repeats domain placed at the N-terminus. A dimerization domain in the middle region acts as a ...GEMIN5 is a predominantly cytoplasmic multifunctional protein, known to be involved in recognizing snRNAs through its WD40 repeats domain placed at the N-terminus. A dimerization domain in the middle region acts as a hub for protein–protein interaction, while a non-canonical RNA-binding site is placed towards the C-terminus. The singular organization of structural domains present in GEMIN5 enables this protein to perform multiple functions through its ability to interact with distinct partners, both RNAs and proteins. This protein exerts a different role in translation regulation depending on its physiological state, such that while GEMIN5 down-regulates global RNA translation, the C-terminal half of the protein promotes translation of its mRNA. Additionally, GEMIN5 is responsible for the preferential partitioning of mRNAs into polysomes. Besides selective translation, GEMIN5 forms part of distinct ribonucleoprotein complexes, reflecting the dynamic organization of macromolecular complexes in response to internal and external signals. In accordance with its contribution to fundamental cellular processes, recent reports described clinical loss of function mutants suggesting that GEMIN5 deficiency is detrimental to cell growth and survival. Remarkably, patients carrying GEMIN5 biallelic variants suffer from neurodevelopmental delay, hypotonia, and cerebellar ataxia. Molecular analyses of individual variants, which are defective in protein dimerization, display decreased levels of ribosome association, reinforcing the involvement of the protein in translation regulation. Importantly, the number of clinical variants and the phenotypic spectrum associated with GEMIN5 disorders is increasing as the knowledge of the protein functions and the pathways linked to its activity augments. Here we discuss relevant advances concerning the functional and structural features of GEMIN5 and its separate domains in RNA-binding, protein interactome, and translation regulation, and how these data can help to understand the involvement of protein malfunction in clinical variants found in patients developing neurodevelopmental disorders.展开更多
Vulnerabilities are a known problem in modern Open Source Software(OSS).Most developers often rely on third-party libraries to accelerate feature implementation.However,these libraries may contain vulnerabilities that...Vulnerabilities are a known problem in modern Open Source Software(OSS).Most developers often rely on third-party libraries to accelerate feature implementation.However,these libraries may contain vulnerabilities that attackers can exploit to propagate malicious code,posing security risks to dependent projects.Existing research addresses these challenges through Software Composition Analysis(SCA)for vulnerability detection and remediation.Nevertheless,current solutions may introduce additional issues,such as incompatibilities,dependency conflicts,and additional vulnerabilities.To address this,we propose Vulnerability Scan and Protection(VulnScanPro),a robust solution for detection and remediation vulnerabilities in Java projects.Specifically,VulnScanPro builds a finegrained method graph to identify unreachable methods.The method graph is mapped to the project’s dependency tree,constructing a comprehensive vulnerability propagation graph that identifies unreachable vulnerable APIs and dependencies.Based on this analysis,we propose three solutions for vulnerability remediation:(1)Removing unreachable vulnerable dependencies,thereby resolving security risks and reducing maintenance overhead.(2)Upgrading vulnerable dependencies to the closest non-vulnerable versions,while pinning the versions of transitive dependencies introduced by the vulnerable dependency,in order to mitigate compatibility issues and prevent the introduction of new vulnerabilities.(3)Eliminating unreachable vulnerable APIs,particularly when security patches are either incompatible or absent.Experimental results show that these solutions effectively mitigate vulnerabilities and enhance the overall security of the project.展开更多
Human motion modeling is a core technology in computer animation,game development,and humancomputer interaction.In particular,generating natural and coherent in-between motion using only the initial and terminal frame...Human motion modeling is a core technology in computer animation,game development,and humancomputer interaction.In particular,generating natural and coherent in-between motion using only the initial and terminal frames remains a fundamental yet unresolved challenge.Existing methods typically rely on dense keyframe inputs or complex prior structures,making it difficult to balance motion quality and plausibility under conditions such as sparse constraints,long-term dependencies,and diverse motion styles.To address this,we propose a motion generation framework based on a frequency-domain diffusion model,which aims to better model complex motion distributions and enhance generation stability under sparse conditions.Our method maps motion sequences to the frequency domain via the Discrete Cosine Transform(DCT),enabling more effective modeling of low-frequency motion structures while suppressing high-frequency noise.A denoising network based on self-attention is introduced to capture long-range temporal dependencies and improve global structural awareness.Additionally,a multi-objective loss function is employed to jointly optimize motion smoothness,pose diversity,and anatomical consistency,enhancing the realism and physical plausibility of the generated sequences.Comparative experiments on the Human3.6M and LaFAN1 datasets demonstrate that our method outperforms state-of-the-art approaches across multiple performance metrics,showing stronger capabilities in generating intermediate motion frames.This research offers a new perspective and methodology for human motion generation and holds promise for applications in character animation,game development,and virtual interaction.展开更多
The next-generation RAN,known as Open Radio Access Network(ORAN),allows for several advantages,including cost-effectiveness,network flexibility,and interoperability.Now ORAN applications,utilising machine learning(ML)...The next-generation RAN,known as Open Radio Access Network(ORAN),allows for several advantages,including cost-effectiveness,network flexibility,and interoperability.Now ORAN applications,utilising machine learning(ML)and artificial intelligence(AI)techniques,have become standard practice.The need for Federated Learning(FL)for ML model training in ORAN environments is heightened by the modularised structure of the ORAN architecture and the shortcomings of conventional ML techniques.However,the traditional plaintext model update sharing of FL in multi-BS contexts is susceptible to privacy violations such as deep-leakage gradient assaults and inference.Therefore,this research presents a novel blockchain-assisted improved cryptographic privacy-preserving federated learning(BICPPFL)model,with the help of ORAN,to safely carry out federated learning and protect privacy.This model improves on the conventional masking technique for sharing model parameters by adding new characteristics.These features include the choice of distributed aggregators,validation for final model aggregation,and individual validation for BSs.To manage the security and privacy of FL processes,a combined homomorphic proxy-reencryption(HPReE)and lattice-cryptographic method(HPReEL)has been used.The upgraded delegated proof of stake(Up-DPoS)consensus protocol,which will provide quick validation of model exchanges and protect against malicious attacks,is employed for effective consensus across blockchain nodes.Without sacrificing performance metrics,the BICPPFL model strengthens privacy and adds security layers while facilitating the transfer of sensitive data across several BSs.The framework is deployed on top of a Hyperledger Fabric blockchain to evaluate its effectiveness.The experimental findings prove the reliability and privacy-preserving capability of the BICPPFL model.展开更多
通过文本分析法和网络调研法,发掘美国科学公共图书馆(Public Library of Science,PLOS)开放科学服务的路径与运作模式,讨论PLOS开放科学服务的特点和发展模式。建议我国依据发展趋势、制定符合国情和科技发展水平的开放科学政策、实现...通过文本分析法和网络调研法,发掘美国科学公共图书馆(Public Library of Science,PLOS)开放科学服务的路径与运作模式,讨论PLOS开放科学服务的特点和发展模式。建议我国依据发展趋势、制定符合国情和科技发展水平的开放科学政策、实现路径、机制、工作模式以及学术出版模式的创新,推动新兴技术、工具在开放科学的应用,提高出版平台对发布成果的及时评价反馈和宣传推广,为我国全面融入开放科学全过程、全方位提升服务能力提供保障。展开更多
文摘为解决由于无人机视角下毛竹林的形状和纹理复杂,现有方法在分割精度和鲁棒性方面表现不佳的问题,提出了一种应用跨领域适应和偏移量引导的毛竹林分割网络——BFSNet。以百山祖国家公园为试验区,利用无人机拍摄周边毛竹林图像构建数据集。为增强模型的特征提取能力,提出跨领域适应模块以有效利用源模型的强特征提取能力,并结合自主学习提取适用于毛竹林分割任务的特征,利用两者的优势进行互补。为提高模型对于不同形状毛竹林的识别和定位能力,结合可变形卷积的偏移量引导模块,引入可学习的偏移量参数,以适应不同形状的毛竹林目标。将BFSNet在DeepGlobe Land Cover Classification Challenge和自制数据集上进行模型训练和测试,并与多种主流图像分割方法进行对比。结果表明:BFSNet在交并比、Dice系数、精确率和召回率4项指标上均取得了最优的性能表现,分别获得了76.04%和71.93%的交并比。与多种主流的图像分割模型相比,BFSNet在毛竹林的分割效果方面表现最为出色,对毛竹林形状的精确建模能力能够有效地应对不同形态的毛竹林。
基金supported by project ZR2022MF330 supported by Shandong Provincial Natural Science Foundationthe National Natural Science Foundation of China under Grant No.61701286.
文摘Synthetic speech detection is an essential task in the field of voice security,aimed at identifying deceptive voice attacks generated by text-to-speech(TTS)systems or voice conversion(VC)systems.In this paper,we propose a synthetic speech detection model called TFTransformer,which integrates both local and global features to enhance detection capabilities by effectively modeling local and global dependencies.Structurally,the model is divided into two main components:a front-end and a back-end.The front-end of the model uses a combination of SincLayer and two-dimensional(2D)convolution to extract high-level feature maps(HFM)containing local dependency of the input speech signals.The back-end uses time-frequency Transformer module to process these feature maps and further capture global dependency.Furthermore,we propose TFTransformer-SE,which incorporates a channel attention mechanism within the 2D convolutional blocks.This enhancement aims to more effectively capture local dependencies,thereby improving the model’s performance.The experiments were conducted on the ASVspoof 2021 LA dataset,and the results showed that the model achieved an equal error rate(EER)of 3.37%without data augmentation.Additionally,we evaluated the model using the ASVspoof 2019 LA dataset,achieving an EER of 0.84%,also without data augmentation.This demonstrates that combining local and global dependencies in the time-frequency domain can significantly improve detection accuracy.
基金partially supported by grants PID2020-115096RB-I00 and PID2023-148273NB-I00 from Ministerio de Ciencia y Universidad (MICIU/AEI)(to EMS)。
文摘GEMIN5 is a predominantly cytoplasmic multifunctional protein, known to be involved in recognizing snRNAs through its WD40 repeats domain placed at the N-terminus. A dimerization domain in the middle region acts as a hub for protein–protein interaction, while a non-canonical RNA-binding site is placed towards the C-terminus. The singular organization of structural domains present in GEMIN5 enables this protein to perform multiple functions through its ability to interact with distinct partners, both RNAs and proteins. This protein exerts a different role in translation regulation depending on its physiological state, such that while GEMIN5 down-regulates global RNA translation, the C-terminal half of the protein promotes translation of its mRNA. Additionally, GEMIN5 is responsible for the preferential partitioning of mRNAs into polysomes. Besides selective translation, GEMIN5 forms part of distinct ribonucleoprotein complexes, reflecting the dynamic organization of macromolecular complexes in response to internal and external signals. In accordance with its contribution to fundamental cellular processes, recent reports described clinical loss of function mutants suggesting that GEMIN5 deficiency is detrimental to cell growth and survival. Remarkably, patients carrying GEMIN5 biallelic variants suffer from neurodevelopmental delay, hypotonia, and cerebellar ataxia. Molecular analyses of individual variants, which are defective in protein dimerization, display decreased levels of ribosome association, reinforcing the involvement of the protein in translation regulation. Importantly, the number of clinical variants and the phenotypic spectrum associated with GEMIN5 disorders is increasing as the knowledge of the protein functions and the pathways linked to its activity augments. Here we discuss relevant advances concerning the functional and structural features of GEMIN5 and its separate domains in RNA-binding, protein interactome, and translation regulation, and how these data can help to understand the involvement of protein malfunction in clinical variants found in patients developing neurodevelopmental disorders.
基金supported by the National Natural Science Foundation of China(Grant No.62141210)the Fundamental Research Funds for the Central Universities(Grant No.N2217005)+1 种基金Open Fund of State Key Lab.for Novel Software Technology,Nanjing University(KFKT2021B01)111 Project(B16009).
文摘Vulnerabilities are a known problem in modern Open Source Software(OSS).Most developers often rely on third-party libraries to accelerate feature implementation.However,these libraries may contain vulnerabilities that attackers can exploit to propagate malicious code,posing security risks to dependent projects.Existing research addresses these challenges through Software Composition Analysis(SCA)for vulnerability detection and remediation.Nevertheless,current solutions may introduce additional issues,such as incompatibilities,dependency conflicts,and additional vulnerabilities.To address this,we propose Vulnerability Scan and Protection(VulnScanPro),a robust solution for detection and remediation vulnerabilities in Java projects.Specifically,VulnScanPro builds a finegrained method graph to identify unreachable methods.The method graph is mapped to the project’s dependency tree,constructing a comprehensive vulnerability propagation graph that identifies unreachable vulnerable APIs and dependencies.Based on this analysis,we propose three solutions for vulnerability remediation:(1)Removing unreachable vulnerable dependencies,thereby resolving security risks and reducing maintenance overhead.(2)Upgrading vulnerable dependencies to the closest non-vulnerable versions,while pinning the versions of transitive dependencies introduced by the vulnerable dependency,in order to mitigate compatibility issues and prevent the introduction of new vulnerabilities.(3)Eliminating unreachable vulnerable APIs,particularly when security patches are either incompatible or absent.Experimental results show that these solutions effectively mitigate vulnerabilities and enhance the overall security of the project.
基金supported by the National Natural Science Foundation of China(Grant No.72161034).
文摘Human motion modeling is a core technology in computer animation,game development,and humancomputer interaction.In particular,generating natural and coherent in-between motion using only the initial and terminal frames remains a fundamental yet unresolved challenge.Existing methods typically rely on dense keyframe inputs or complex prior structures,making it difficult to balance motion quality and plausibility under conditions such as sparse constraints,long-term dependencies,and diverse motion styles.To address this,we propose a motion generation framework based on a frequency-domain diffusion model,which aims to better model complex motion distributions and enhance generation stability under sparse conditions.Our method maps motion sequences to the frequency domain via the Discrete Cosine Transform(DCT),enabling more effective modeling of low-frequency motion structures while suppressing high-frequency noise.A denoising network based on self-attention is introduced to capture long-range temporal dependencies and improve global structural awareness.Additionally,a multi-objective loss function is employed to jointly optimize motion smoothness,pose diversity,and anatomical consistency,enhancing the realism and physical plausibility of the generated sequences.Comparative experiments on the Human3.6M and LaFAN1 datasets demonstrate that our method outperforms state-of-the-art approaches across multiple performance metrics,showing stronger capabilities in generating intermediate motion frames.This research offers a new perspective and methodology for human motion generation and holds promise for applications in character animation,game development,and virtual interaction.
文摘The next-generation RAN,known as Open Radio Access Network(ORAN),allows for several advantages,including cost-effectiveness,network flexibility,and interoperability.Now ORAN applications,utilising machine learning(ML)and artificial intelligence(AI)techniques,have become standard practice.The need for Federated Learning(FL)for ML model training in ORAN environments is heightened by the modularised structure of the ORAN architecture and the shortcomings of conventional ML techniques.However,the traditional plaintext model update sharing of FL in multi-BS contexts is susceptible to privacy violations such as deep-leakage gradient assaults and inference.Therefore,this research presents a novel blockchain-assisted improved cryptographic privacy-preserving federated learning(BICPPFL)model,with the help of ORAN,to safely carry out federated learning and protect privacy.This model improves on the conventional masking technique for sharing model parameters by adding new characteristics.These features include the choice of distributed aggregators,validation for final model aggregation,and individual validation for BSs.To manage the security and privacy of FL processes,a combined homomorphic proxy-reencryption(HPReE)and lattice-cryptographic method(HPReEL)has been used.The upgraded delegated proof of stake(Up-DPoS)consensus protocol,which will provide quick validation of model exchanges and protect against malicious attacks,is employed for effective consensus across blockchain nodes.Without sacrificing performance metrics,the BICPPFL model strengthens privacy and adds security layers while facilitating the transfer of sensitive data across several BSs.The framework is deployed on top of a Hyperledger Fabric blockchain to evaluate its effectiveness.The experimental findings prove the reliability and privacy-preserving capability of the BICPPFL model.
文摘通过文本分析法和网络调研法,发掘美国科学公共图书馆(Public Library of Science,PLOS)开放科学服务的路径与运作模式,讨论PLOS开放科学服务的特点和发展模式。建议我国依据发展趋势、制定符合国情和科技发展水平的开放科学政策、实现路径、机制、工作模式以及学术出版模式的创新,推动新兴技术、工具在开放科学的应用,提高出版平台对发布成果的及时评价反馈和宣传推广,为我国全面融入开放科学全过程、全方位提升服务能力提供保障。
