The three-dimensional weight function method recently developed by the authors is used to determine stress intensity factors for two symmetric quarter-elliptical corner cracks at a hole in a wide finite-thickness plat...The three-dimensional weight function method recently developed by the authors is used to determine stress intensity factors for two symmetric quarter-elliptical corner cracks at a hole in a wide finite-thickness plate subjected to remote tensile loading. The geometry parameters considered are r / t = 0.5, 1, 2; a / c= 0.2, 0.5, 1, 2; a / t = 0.2, 0.5 within c/r= 2. The results are compared, where possible, with other solutions available in the literature. Generally good agreement is observed. The effect of an approximation of the two-dimensional unflawed stress distribution on the accuracy of stress intensity factors by the weight function method is discussed.展开更多
自然场景下的文本检测任务是图像处理领域中的难点之一.EAST(Efficient and Accurate Scene Text detector)算法是近年来比较出色的文本检测算法,但是增加后置处理之后的AdvancedEAST算法仍存在由于激活像素的头尾边界丢失导致的漏检情...自然场景下的文本检测任务是图像处理领域中的难点之一.EAST(Efficient and Accurate Scene Text detector)算法是近年来比较出色的文本检测算法,但是增加后置处理之后的AdvancedEAST算法仍存在由于激活像素的头尾边界丢失导致的漏检情况,对密集文本的检测效果也不是很理想.因此提出了Dilated-Corner Attention EAST(DCA_EAST)改进算法,对网络结构加入空洞卷积模块以及角点注意力模块,改善了漏检情况.针对损失函数,加入类别权重因子和样本难度权重因子,有效提升了密集文本的检测效果.实验结果表明,该算法在ICDAR2019的ReCTS数据集上准确率为93.02%,召回率为76.69%,F-measured值为84.07%,优于AdvancedEAST算法.展开更多
文摘The three-dimensional weight function method recently developed by the authors is used to determine stress intensity factors for two symmetric quarter-elliptical corner cracks at a hole in a wide finite-thickness plate subjected to remote tensile loading. The geometry parameters considered are r / t = 0.5, 1, 2; a / c= 0.2, 0.5, 1, 2; a / t = 0.2, 0.5 within c/r= 2. The results are compared, where possible, with other solutions available in the literature. Generally good agreement is observed. The effect of an approximation of the two-dimensional unflawed stress distribution on the accuracy of stress intensity factors by the weight function method is discussed.
文摘自然场景下的文本检测任务是图像处理领域中的难点之一.EAST(Efficient and Accurate Scene Text detector)算法是近年来比较出色的文本检测算法,但是增加后置处理之后的AdvancedEAST算法仍存在由于激活像素的头尾边界丢失导致的漏检情况,对密集文本的检测效果也不是很理想.因此提出了Dilated-Corner Attention EAST(DCA_EAST)改进算法,对网络结构加入空洞卷积模块以及角点注意力模块,改善了漏检情况.针对损失函数,加入类别权重因子和样本难度权重因子,有效提升了密集文本的检测效果.实验结果表明,该算法在ICDAR2019的ReCTS数据集上准确率为93.02%,召回率为76.69%,F-measured值为84.07%,优于AdvancedEAST算法.
