Dengue is one of the most prominent tropical epidemic diseases present in the Rio de Janeiro city and Southeast part of Brazil, due to the widespread conditions of occurrence of the dengue vector, the mosquito Aedesae...Dengue is one of the most prominent tropical epidemic diseases present in the Rio de Janeiro city and Southeast part of Brazil, due to the widespread conditions of occurrence of the dengue vector, the mosquito Aedesaegypti, such as high-temperature days interlaced with afternoon or nocturnal rainstorms in summer. This work has the objective of investigating the relationships between variabilities of the El Ni?o-South Oscillation (ENSO) and greater epidemics of dengue in Rio de Janeiro city. To accomplish this goal, the analysis and signal decomposition by cross-wavelet transform (WT) was applied to obtain the cross variability associated with variations of power and phase of both signals by characteristic periods and along with the time series. Data considered in the analysis are (the decimal logarithm of normalized value) of the monthly available notifications of dengue worsening, provided by the public health system of Brazil, and the Southern Oscillation Index (SOI) Ni?o 3.4 data, provided by the National Oceanic and Atmospheric Administration (NOAA), in the period 2000-2017. A maximum cross-wavelet power close to 0.45 was obtained for the representative period of 1 year and also to periods between 3 and 4 years, associated with the positive phase of the SOI index (i.e. , La Ni?a) or with a transition to the positive phase. The evolution of the combined variability of SOI and dengue can be expressed by progressive differences in phase along the time, eventually resulting in yielding phases (i.e., La Niña-Dengue epidemic).展开更多
针对主流Transformer网络仅对输入像素块做自注意力计算而忽略了不同像素块间的信息交互,以及输入尺度单一导致局部特征细节模糊的问题,本文提出一种基于Transformer并用于处理视觉任务的主干网络ConvFormer. ConvFormer通过所设计的多...针对主流Transformer网络仅对输入像素块做自注意力计算而忽略了不同像素块间的信息交互,以及输入尺度单一导致局部特征细节模糊的问题,本文提出一种基于Transformer并用于处理视觉任务的主干网络ConvFormer. ConvFormer通过所设计的多尺度混洗自注意力模块(Channel-Shuffle and Multi-Scale attention,CSMS)和动态相对位置编码模块(Dynamic Relative Position Coding,DRPC)来聚合多尺度像素块间的语义信息,并在前馈网络中引入深度卷积提高网络的局部建模能力.在公开数据集ImageNet-1K,COCO 2017和ADE20K上分别进行图像分类、目标检测和语义分割实验,ConvFormer-Tiny与不同视觉任务中同量级最优网络RetNetY-4G,Swin-Tiny和ResNet50对比,精度分别提高0.3%,1.4%和0.5%.展开更多
遥感图像的道路分割任务是遥感应用领域的一个研究热点,一直受到广泛的关注。由于遥感图像天然具备背景复杂、目标密集等特性,全局语义信息的构建对于准确提取遥感图像中道路是至关重要的。因此,基于Transformer模型进行优化,提出了基...遥感图像的道路分割任务是遥感应用领域的一个研究热点,一直受到广泛的关注。由于遥感图像天然具备背景复杂、目标密集等特性,全局语义信息的构建对于准确提取遥感图像中道路是至关重要的。因此,基于Transformer模型进行优化,提出了基于空间可分离注意力的跨尺度令牌嵌入Transformer遥感道路提取模型Cross-RoadFormer。具体而言,针对图像中道路尺度不统一的问题,设计了跨尺度编码层,将不同尺度的特征编码作为一个令牌嵌入整体,作为Transformer的输入,解决了Transformer跨尺度交互的问题;此外,提出了一种空间可分离注意力,其中,局部分组注意力获取细粒度、短距离信息,全局采样注意力捕获长距离、全局上下文信息,在保证道路提取准确度的前提下,降低了模型的计算量。在Massachusetts数据集和DeepGlobe数据集上的实验表明,提出的Cross-RoadFormer都实现了更高的IoU(intersection over union),分别为68.40%和58.04%,展现了该方法的优越性。展开更多
文摘Dengue is one of the most prominent tropical epidemic diseases present in the Rio de Janeiro city and Southeast part of Brazil, due to the widespread conditions of occurrence of the dengue vector, the mosquito Aedesaegypti, such as high-temperature days interlaced with afternoon or nocturnal rainstorms in summer. This work has the objective of investigating the relationships between variabilities of the El Ni?o-South Oscillation (ENSO) and greater epidemics of dengue in Rio de Janeiro city. To accomplish this goal, the analysis and signal decomposition by cross-wavelet transform (WT) was applied to obtain the cross variability associated with variations of power and phase of both signals by characteristic periods and along with the time series. Data considered in the analysis are (the decimal logarithm of normalized value) of the monthly available notifications of dengue worsening, provided by the public health system of Brazil, and the Southern Oscillation Index (SOI) Ni?o 3.4 data, provided by the National Oceanic and Atmospheric Administration (NOAA), in the period 2000-2017. A maximum cross-wavelet power close to 0.45 was obtained for the representative period of 1 year and also to periods between 3 and 4 years, associated with the positive phase of the SOI index (i.e. , La Ni?a) or with a transition to the positive phase. The evolution of the combined variability of SOI and dengue can be expressed by progressive differences in phase along the time, eventually resulting in yielding phases (i.e., La Niña-Dengue epidemic).
文摘针对主流Transformer网络仅对输入像素块做自注意力计算而忽略了不同像素块间的信息交互,以及输入尺度单一导致局部特征细节模糊的问题,本文提出一种基于Transformer并用于处理视觉任务的主干网络ConvFormer. ConvFormer通过所设计的多尺度混洗自注意力模块(Channel-Shuffle and Multi-Scale attention,CSMS)和动态相对位置编码模块(Dynamic Relative Position Coding,DRPC)来聚合多尺度像素块间的语义信息,并在前馈网络中引入深度卷积提高网络的局部建模能力.在公开数据集ImageNet-1K,COCO 2017和ADE20K上分别进行图像分类、目标检测和语义分割实验,ConvFormer-Tiny与不同视觉任务中同量级最优网络RetNetY-4G,Swin-Tiny和ResNet50对比,精度分别提高0.3%,1.4%和0.5%.
文摘遥感图像的道路分割任务是遥感应用领域的一个研究热点,一直受到广泛的关注。由于遥感图像天然具备背景复杂、目标密集等特性,全局语义信息的构建对于准确提取遥感图像中道路是至关重要的。因此,基于Transformer模型进行优化,提出了基于空间可分离注意力的跨尺度令牌嵌入Transformer遥感道路提取模型Cross-RoadFormer。具体而言,针对图像中道路尺度不统一的问题,设计了跨尺度编码层,将不同尺度的特征编码作为一个令牌嵌入整体,作为Transformer的输入,解决了Transformer跨尺度交互的问题;此外,提出了一种空间可分离注意力,其中,局部分组注意力获取细粒度、短距离信息,全局采样注意力捕获长距离、全局上下文信息,在保证道路提取准确度的前提下,降低了模型的计算量。在Massachusetts数据集和DeepGlobe数据集上的实验表明,提出的Cross-RoadFormer都实现了更高的IoU(intersection over union),分别为68.40%和58.04%,展现了该方法的优越性。
