针对跌倒检测任务中复杂信息干扰和数据集缺乏导致模型精度不高的问题,设计一种高精度跌倒检测算法,降低模型参数的同时保持各种场景下的鲁棒性。该算法基于YOLOv5s改进,在骨干网络中使用Ghost module和解耦全连接注意力,以较低计算开...针对跌倒检测任务中复杂信息干扰和数据集缺乏导致模型精度不高的问题,设计一种高精度跌倒检测算法,降低模型参数的同时保持各种场景下的鲁棒性。该算法基于YOLOv5s改进,在骨干网络中使用Ghost module和解耦全连接注意力,以较低计算开销提升模型在光线变化、遮挡等干扰环境下的性能。在颈部层使用自适应感受野和空间通道混合注意力,提升神经元对不同尺度特征的适应性,应对人体形变、视角变化等干扰。引入EIoU损失函数,加速收敛提升训练精度。在公开数据集Le2i Fall Detection Dataset和UR Fall Detection上,精确率、召回率、mAP0.5和mAP(0.5:0.95)相比YOLOv5s分别提高4.0%,4.2%,2.9%和4.3%,参数量降低38.6%。该算法在多种应用场景下都保持较高检测精度,参数量较低,满足实际应用场景部署要求。展开更多
针对复杂背景下红外场景对比度低、特征不足、细节不清而导致的目标检测效率低的问题,在YOLOv5s模型基础上通过创建TCC(two-way convolution and Concat)模块并引入华为Ghost模块,提出了一种基于改进YOLOv5s模型的红外弱小目标检测方法...针对复杂背景下红外场景对比度低、特征不足、细节不清而导致的目标检测效率低的问题,在YOLOv5s模型基础上通过创建TCC(two-way convolution and Concat)模块并引入华为Ghost模块,提出了一种基于改进YOLOv5s模型的红外弱小目标检测方法。首先,结合红外图像的低级语义特征,采取二路卷积和多尺度思想创建了TCC模块,提升了特征提取的全面性;接着,为进一步简化网络结构、减少网络参数量,引入轻量化Ghost模块改进了SPP池化层和CSP2卷积网络;最后,以无人机为实验对象,构建了白天和夜间不同背景条件下的红外弱小目标数据集,实验验证了本文改进算法的有效性。结果表明:改进后的YOLOv5s模型在较少损失帧频的情况下,检测精度提升了1.34%,平均精度均值(mean average precision, mAP)提升了2.26%,优于YOLOv4-tiny和YOLOv7-tiny两种轻量化模型,并与YOLOv8s模型精度相当,但模型参数量仅为YOLOv8s模型的53%,完全可以满足嵌入式设备部署的需求。展开更多
为了降低服装目标检测模型的参数量和浮点型计算量,提出一种改进的轻量级服装目标检测模型——GYOLOv5s.首先使用Ghost卷积重构YOLOv5s的主干网络;然后使用DeepFashion2数据集中的部分数据进行模型训练和验证;最后将训练好的模型用于服...为了降低服装目标检测模型的参数量和浮点型计算量,提出一种改进的轻量级服装目标检测模型——GYOLOv5s.首先使用Ghost卷积重构YOLOv5s的主干网络;然后使用DeepFashion2数据集中的部分数据进行模型训练和验证;最后将训练好的模型用于服装图像的目标检测.实验结果表明,G-YOLOv5s的mAP达到71.7%,模型体积为9.09 MB,浮点型计算量为9.8 G FLOPs,与改进前的YOLOv5s网络相比,模型体积压缩了34.8%,计算量减少了41.3%,精度仅下降1.3%,方便部署在资源有限的设备中使用.展开更多
Since the publication of Coraz6n tan blanco （A Heart so White） （1992）, many critics have compared the Spanish novelist Javier Marias to Marcel Proust. Both favor long, meandering sentences, in which they insert vo...Since the publication of Coraz6n tan blanco （A Heart so White） （1992）, many critics have compared the Spanish novelist Javier Marias to Marcel Proust. Both favor long, meandering sentences, in which they insert voluminous asides. In thematic terms, their narratives are constantly involved with meditation over the extent to which we can understand the past, or the degree to which we can know either ourselves or others. Beyond their common preoccupation with time and memory, I will consider some remarkable similarities between Marias＇ and Proust＇s formative years and the role translation played in the development of their style. I will show the many ways in which Proust ＂haunts＂ Marias： in his metaphorical use of the translating practice, in his love of deferral, and in his brooding first-person narrators, racked by the anxiety of ignorance.展开更多
This article aims to present Beloved, a ghost-in-the-flesh protagonist of Toni Morrison＇s Pulitzer Prize-winning novel Beloved （1987）, as an incarnation of memory of slavery. Interpreted as personal and shared expe...This article aims to present Beloved, a ghost-in-the-flesh protagonist of Toni Morrison＇s Pulitzer Prize-winning novel Beloved （1987）, as an incarnation of memory of slavery. Interpreted as personal and shared experience, Beloved will be analyzed on the basis of memory＇s dynamic nature as an active mnemonic agent operating in and between the individual and collective zones. It will be also argued that on the one hand, Beloved embodies memories of past slaved lives of the novel＇s central characters, Sethe and Paul D, while on the other hand she plays the role of an allegoric reminder of all Black slaves who lived and died in bondage on the American continent. Finally, Beloved will be symbolically seen as a historical, cultural and psychological link between contemporary African Americans and their African ancestors of the Middle Passage. The theoretical framework for this study of Morrison＇s most memorable ghost figure will follow from a discussion of memory＇s individual and shared qualities, as well as from the concepts of a collective consciousness, the collective unconscious, and collective memory.展开更多
文摘针对跌倒检测任务中复杂信息干扰和数据集缺乏导致模型精度不高的问题,设计一种高精度跌倒检测算法,降低模型参数的同时保持各种场景下的鲁棒性。该算法基于YOLOv5s改进,在骨干网络中使用Ghost module和解耦全连接注意力,以较低计算开销提升模型在光线变化、遮挡等干扰环境下的性能。在颈部层使用自适应感受野和空间通道混合注意力,提升神经元对不同尺度特征的适应性,应对人体形变、视角变化等干扰。引入EIoU损失函数,加速收敛提升训练精度。在公开数据集Le2i Fall Detection Dataset和UR Fall Detection上,精确率、召回率、mAP0.5和mAP(0.5:0.95)相比YOLOv5s分别提高4.0%,4.2%,2.9%和4.3%,参数量降低38.6%。该算法在多种应用场景下都保持较高检测精度,参数量较低,满足实际应用场景部署要求。
文摘针对复杂背景下红外场景对比度低、特征不足、细节不清而导致的目标检测效率低的问题,在YOLOv5s模型基础上通过创建TCC(two-way convolution and Concat)模块并引入华为Ghost模块,提出了一种基于改进YOLOv5s模型的红外弱小目标检测方法。首先,结合红外图像的低级语义特征,采取二路卷积和多尺度思想创建了TCC模块,提升了特征提取的全面性;接着,为进一步简化网络结构、减少网络参数量,引入轻量化Ghost模块改进了SPP池化层和CSP2卷积网络;最后,以无人机为实验对象,构建了白天和夜间不同背景条件下的红外弱小目标数据集,实验验证了本文改进算法的有效性。结果表明:改进后的YOLOv5s模型在较少损失帧频的情况下,检测精度提升了1.34%,平均精度均值(mean average precision, mAP)提升了2.26%,优于YOLOv4-tiny和YOLOv7-tiny两种轻量化模型,并与YOLOv8s模型精度相当,但模型参数量仅为YOLOv8s模型的53%,完全可以满足嵌入式设备部署的需求。
文摘为了降低服装目标检测模型的参数量和浮点型计算量,提出一种改进的轻量级服装目标检测模型——GYOLOv5s.首先使用Ghost卷积重构YOLOv5s的主干网络;然后使用DeepFashion2数据集中的部分数据进行模型训练和验证;最后将训练好的模型用于服装图像的目标检测.实验结果表明,G-YOLOv5s的mAP达到71.7%,模型体积为9.09 MB,浮点型计算量为9.8 G FLOPs,与改进前的YOLOv5s网络相比,模型体积压缩了34.8%,计算量减少了41.3%,精度仅下降1.3%,方便部署在资源有限的设备中使用.
文摘Since the publication of Coraz6n tan blanco （A Heart so White） （1992）, many critics have compared the Spanish novelist Javier Marias to Marcel Proust. Both favor long, meandering sentences, in which they insert voluminous asides. In thematic terms, their narratives are constantly involved with meditation over the extent to which we can understand the past, or the degree to which we can know either ourselves or others. Beyond their common preoccupation with time and memory, I will consider some remarkable similarities between Marias＇ and Proust＇s formative years and the role translation played in the development of their style. I will show the many ways in which Proust ＂haunts＂ Marias： in his metaphorical use of the translating practice, in his love of deferral, and in his brooding first-person narrators, racked by the anxiety of ignorance.
文摘This article aims to present Beloved, a ghost-in-the-flesh protagonist of Toni Morrison＇s Pulitzer Prize-winning novel Beloved （1987）, as an incarnation of memory of slavery. Interpreted as personal and shared experience, Beloved will be analyzed on the basis of memory＇s dynamic nature as an active mnemonic agent operating in and between the individual and collective zones. It will be also argued that on the one hand, Beloved embodies memories of past slaved lives of the novel＇s central characters, Sethe and Paul D, while on the other hand she plays the role of an allegoric reminder of all Black slaves who lived and died in bondage on the American continent. Finally, Beloved will be symbolically seen as a historical, cultural and psychological link between contemporary African Americans and their African ancestors of the Middle Passage. The theoretical framework for this study of Morrison＇s most memorable ghost figure will follow from a discussion of memory＇s individual and shared qualities, as well as from the concepts of a collective consciousness, the collective unconscious, and collective memory.
