针对基于集合预测的密集视频描述方法由于缺乏显式的事件间特征交互且未针对事件间差异训练模型而导致的模型重复预测事件或生成语句雷同问题,提出一种基于事件最大边界的密集视频描述(dense video captioning based on event maximal m...针对基于集合预测的密集视频描述方法由于缺乏显式的事件间特征交互且未针对事件间差异训练模型而导致的模型重复预测事件或生成语句雷同问题,提出一种基于事件最大边界的密集视频描述(dense video captioning based on event maximal margin,EMM-DVC)方法。事件边界是包含事件间特征相似度、事件在视频中时间位置的距离、生成描述多样性的评分。EMM-DVC通过最大化事件边界,使相似预测结果的距离远且预测结果和实际事件的距离近。另外,EMM-DVC引入事件边界距离损失函数,通过扩大事件边界距离,引导模型关注不同事件。在ActivityNet Captions数据集上的实验证明,EMM-DVC与同类密集视频描述模型相比能生成更具多样性的描述文本,并且与主流密集视频描述模型相比,EMM-DVC在多个指标上达到最优水平。展开更多
For electromagnetic scattering of 3?D complex electrically large conducting targets,a new hybrid algorithm,MoM?PO/SBR algorithm,is presented to realize the interaction of information between method of moment(MoM)and p...For electromagnetic scattering of 3?D complex electrically large conducting targets,a new hybrid algorithm,MoM?PO/SBR algorithm,is presented to realize the interaction of information between method of moment(MoM)and physical optics(PO)/shooting and bouncing ray(SBR).In the algorithm,the COC file that based on the Huygens equivalent principle is introduced,and the conversion interface between the equivalent surface and the target is established.And then,the multi?task flow model presented in this paper is adopted to conduct CPU/graphics processing unit(GPU)tests of the algorithm under three modes,i.e.,MPI/OpenMP,MPI/compute unified device architecture(CUDA)and multi?task programming model(MTPM).Numerical results are presented and compared with reference solutions in order to illustrate the accuracy and the efficiency of the proposed algorithm.展开更多
A real-time operating system (RTOS), also named OS, is designed based on the hardware platform of MC68376, and is implemented in the electronic control system for unit pump in diesel engine. A parallel and time-base...A real-time operating system (RTOS), also named OS, is designed based on the hardware platform of MC68376, and is implemented in the electronic control system for unit pump in diesel engine. A parallel and time-based task division method is introduced and the multi-task software architecture is built in the software system for electronic unit pump (EUP) system. The V-model software development process is used to control algorithm of each task. The simulation results of the hardware-in-the-loop simulation system (HILSS) and the engine experimental results show that the OS is an efficient real-time kernel, and can meet the real-time demands of EUP system; The built multi-task software system is real-time, determinate and reliable. V-model development is a good development process of control algorithms for EUP system, the control precision of control system can be ensured, and the development cycle and cost are also decreased.展开更多
多轮对话是人工智能领域的一个重要分支.如何从多轮对话上下文中正确提取与问题相关的核心内容是多轮对话任务的关键问题.现有模型存在辅助任务低效,对全局与局部信息的筛选不够充分,对较短的多轮对话数据学习能力不足等问题.针对上述问...多轮对话是人工智能领域的一个重要分支.如何从多轮对话上下文中正确提取与问题相关的核心内容是多轮对话任务的关键问题.现有模型存在辅助任务低效,对全局与局部信息的筛选不够充分,对较短的多轮对话数据学习能力不足等问题.针对上述问题,本文提出了一种局部信息增强且能够感知对话结构的多轮对话模型(Structure-aware Dialogue Model with Fine-grained Local Information,SAFL).针对子任务训练代价大的问题,提出了随机滑动窗口回复预测任务,在多轮对话上下文中的不同位置与大小的窗口内进行回复预测,充分学习细粒度的局部对话语义.针对信息筛选不够充分的问题,提出了重点局部信息蒸馏机制,借助多门控融合方法从全局和局部信息之中蒸馏出重点信息,提升模型融合效果.针对模型对较短的多轮对话上下文学习能力不足的问题,提出阶段信息学习机制,在微调前加强预训练语言模型对短多轮对话数据的领域学习,降低微调阶段中对短多轮对话的学习难度.此外,SAFL设计了对话结构感知任务在对话结构方面进一步加强模型对对话上下文的理解能力.Ubuntu和E-commerce数据集上的实验结果表明,SAFL模型的总体性能优于对比模型.展开更多
基于视觉的手部位姿估计技术应用于诸多领域,具备着广泛的国际应用市场前景和巨大发展潜力。然而,手部自身存在检测目标过小、手指高自由度以及手部自遮挡等问题。通过对目前存在的难点分析,将手部位姿估计任务分为手部检测和手部关键...基于视觉的手部位姿估计技术应用于诸多领域,具备着广泛的国际应用市场前景和巨大发展潜力。然而,手部自身存在检测目标过小、手指高自由度以及手部自遮挡等问题。通过对目前存在的难点分析,将手部位姿估计任务分为手部检测和手部关键点检测,提出基于改进的Faster R-CNN的手部位姿估计方法。首先提出基于改进的Faster R-CNN手部检测网络,将传统Faster R-CNN网络中的对ROI(regional of interest)的最大值池化,更改为ROI Align,并增加损失函数用于区分左右手。在此基础上增加了头网络分支用以训练输出MANO(hand model with articulated and non-rigid deformations)手部模型的姿态参数和形状参数,得到手部关键点三维坐标,最终得到手部的三维位姿估计结果。实验表明,手部检测结果中存在的自遮挡和尺度问题得到了解决,并且检测结果的准确性有所提高,本文手部检测算法准确率为85%,比传统Faster R-CNN算法提升13%。手部关键点提取算法在MSRA、ICVL、NYU三个数据集分别取得关键点坐标的均方误差值(key-point mean square error,KMSE)为7.50、6.32、8.50的结果。展开更多
文摘针对基于集合预测的密集视频描述方法由于缺乏显式的事件间特征交互且未针对事件间差异训练模型而导致的模型重复预测事件或生成语句雷同问题,提出一种基于事件最大边界的密集视频描述(dense video captioning based on event maximal margin,EMM-DVC)方法。事件边界是包含事件间特征相似度、事件在视频中时间位置的距离、生成描述多样性的评分。EMM-DVC通过最大化事件边界,使相似预测结果的距离远且预测结果和实际事件的距离近。另外,EMM-DVC引入事件边界距离损失函数,通过扩大事件边界距离,引导模型关注不同事件。在ActivityNet Captions数据集上的实验证明,EMM-DVC与同类密集视频描述模型相比能生成更具多样性的描述文本,并且与主流密集视频描述模型相比,EMM-DVC在多个指标上达到最优水平。
文摘For electromagnetic scattering of 3?D complex electrically large conducting targets,a new hybrid algorithm,MoM?PO/SBR algorithm,is presented to realize the interaction of information between method of moment(MoM)and physical optics(PO)/shooting and bouncing ray(SBR).In the algorithm,the COC file that based on the Huygens equivalent principle is introduced,and the conversion interface between the equivalent surface and the target is established.And then,the multi?task flow model presented in this paper is adopted to conduct CPU/graphics processing unit(GPU)tests of the algorithm under three modes,i.e.,MPI/OpenMP,MPI/compute unified device architecture(CUDA)and multi?task programming model(MTPM).Numerical results are presented and compared with reference solutions in order to illustrate the accuracy and the efficiency of the proposed algorithm.
文摘A real-time operating system (RTOS), also named OS, is designed based on the hardware platform of MC68376, and is implemented in the electronic control system for unit pump in diesel engine. A parallel and time-based task division method is introduced and the multi-task software architecture is built in the software system for electronic unit pump (EUP) system. The V-model software development process is used to control algorithm of each task. The simulation results of the hardware-in-the-loop simulation system (HILSS) and the engine experimental results show that the OS is an efficient real-time kernel, and can meet the real-time demands of EUP system; The built multi-task software system is real-time, determinate and reliable. V-model development is a good development process of control algorithms for EUP system, the control precision of control system can be ensured, and the development cycle and cost are also decreased.
文摘多轮对话是人工智能领域的一个重要分支.如何从多轮对话上下文中正确提取与问题相关的核心内容是多轮对话任务的关键问题.现有模型存在辅助任务低效,对全局与局部信息的筛选不够充分,对较短的多轮对话数据学习能力不足等问题.针对上述问题,本文提出了一种局部信息增强且能够感知对话结构的多轮对话模型(Structure-aware Dialogue Model with Fine-grained Local Information,SAFL).针对子任务训练代价大的问题,提出了随机滑动窗口回复预测任务,在多轮对话上下文中的不同位置与大小的窗口内进行回复预测,充分学习细粒度的局部对话语义.针对信息筛选不够充分的问题,提出了重点局部信息蒸馏机制,借助多门控融合方法从全局和局部信息之中蒸馏出重点信息,提升模型融合效果.针对模型对较短的多轮对话上下文学习能力不足的问题,提出阶段信息学习机制,在微调前加强预训练语言模型对短多轮对话数据的领域学习,降低微调阶段中对短多轮对话的学习难度.此外,SAFL设计了对话结构感知任务在对话结构方面进一步加强模型对对话上下文的理解能力.Ubuntu和E-commerce数据集上的实验结果表明,SAFL模型的总体性能优于对比模型.
文摘基于视觉的手部位姿估计技术应用于诸多领域,具备着广泛的国际应用市场前景和巨大发展潜力。然而,手部自身存在检测目标过小、手指高自由度以及手部自遮挡等问题。通过对目前存在的难点分析,将手部位姿估计任务分为手部检测和手部关键点检测,提出基于改进的Faster R-CNN的手部位姿估计方法。首先提出基于改进的Faster R-CNN手部检测网络,将传统Faster R-CNN网络中的对ROI(regional of interest)的最大值池化,更改为ROI Align,并增加损失函数用于区分左右手。在此基础上增加了头网络分支用以训练输出MANO(hand model with articulated and non-rigid deformations)手部模型的姿态参数和形状参数,得到手部关键点三维坐标,最终得到手部的三维位姿估计结果。实验表明,手部检测结果中存在的自遮挡和尺度问题得到了解决,并且检测结果的准确性有所提高,本文手部检测算法准确率为85%,比传统Faster R-CNN算法提升13%。手部关键点提取算法在MSRA、ICVL、NYU三个数据集分别取得关键点坐标的均方误差值(key-point mean square error,KMSE)为7.50、6.32、8.50的结果。