为了给驾驶员提供实时准确的行人信息、减少交通事故的发生,提出一种检测增强型YOLOv3-tiny(detection of enhanced YOLOv3-tiny,DOEYT)行人检测算法.创建鲁棒的特征提取网络,首先使用非对称最大池化进行下采样,防止随着感受野增大行人...为了给驾驶员提供实时准确的行人信息、减少交通事故的发生,提出一种检测增强型YOLOv3-tiny(detection of enhanced YOLOv3-tiny,DOEYT)行人检测算法.创建鲁棒的特征提取网络,首先使用非对称最大池化进行下采样,防止随着感受野增大行人横向特征的丢失;其次使用Hardswish作为卷积层的激活函数优化网络性能;最后使用GC(globe context)自注意力机制获得全文特征信息.在分类回归网络部分,采用三尺度检测策略,提升小尺度行人目标的检测精度;使用k-means++算法重新生成数据集锚框,提高网络收敛速度.构建行人检测数据集并分为训练集和测试集,对DOEYT算法的性能进行试验验证.结果表明,非对称最大池化、Hardswish函数、GC自注意力机制分别使平均准确率AP提高14.4%、7.9%、10.8%;DOEYT算法在测试集上检测的平均准确率高达91.2%,检测速度为103帧/s,可见该算法可快速准确地检测行人,降低交通事故发生的风险.展开更多
Retinal fundus images are used to discover many diseases.Several Machine learning algorithms are designed to identify the Glaucoma disease.But the accuracy and time consumption performance were not improved.To address...Retinal fundus images are used to discover many diseases.Several Machine learning algorithms are designed to identify the Glaucoma disease.But the accuracy and time consumption performance were not improved.To address this problem Max Pool Convolution Neural Kuan Filtered Tobit Regres-sive Segmentation based Radial Basis Image Classifier(MPCNKFTRS-RBIC)Model is used for detecting the Glaucoma and Stargardt’s disease by early period using higher accuracy and minimal time.In MPCNKFTRS-RBIC Model,the ret-inal fundus image is considered as an input which is preprocessed in hidden layer 1 using weighted adaptive Kuanfilter.Then,preprocessed retinal fundus is given for hidden layer 2 for extracting the features like color,intensity,texture with higher accuracy.After extracting these features,the Tobit Regressive Segmenta-tion process is performed by hidden layer 3 for partitioning preprocessed image within more segments by analyzing the pixel with the extracted features of the fundus image.Then,the segmented image was given to output layer.The radial basis function analyzes the testing image region of a particular class as well as training image region with higher accuracy and minimum time consumption.Simulation is performed with retinal fundus image dataset with various perfor-mance metrics namely peak signal-to-noise ratio,accuracy and time,error rate concerning several retina fundus image and image size.展开更多
高效的图像特征表示是计算机视觉的基础.基于图像的视觉显著性机制及深度学习模型的思想,提出一种融合图像显著性的层次稀疏特征表示用于图像分类.这种层次特征学习每一层都由3个部分组成:稀疏编码、显著性最大值汇聚(saliency max pool...高效的图像特征表示是计算机视觉的基础.基于图像的视觉显著性机制及深度学习模型的思想,提出一种融合图像显著性的层次稀疏特征表示用于图像分类.这种层次特征学习每一层都由3个部分组成:稀疏编码、显著性最大值汇聚(saliency max pooling)和对比度归一化.通过在图像层次稀疏表示中引入图像显著信息,加强了图像特征的语义信息,得到图像显著特征表示.相比于手工指定特征,该模型采用无监督数据驱动的方式直接从图像中学习到有效的图像特征描述.最后采用支持向量机(support vector machine,SVM)分类器进行监督学习,实现对图像进行分类.在2个常用的标准图像数据集(Caltech 101和Caltech 256)上进行的实验结果表明,结合图像显著性信息的层次特征表示,相比于基于局部特征的单层稀疏表示在分类性能上有了显著提升.展开更多
文摘为了给驾驶员提供实时准确的行人信息、减少交通事故的发生,提出一种检测增强型YOLOv3-tiny(detection of enhanced YOLOv3-tiny,DOEYT)行人检测算法.创建鲁棒的特征提取网络,首先使用非对称最大池化进行下采样,防止随着感受野增大行人横向特征的丢失;其次使用Hardswish作为卷积层的激活函数优化网络性能;最后使用GC(globe context)自注意力机制获得全文特征信息.在分类回归网络部分,采用三尺度检测策略,提升小尺度行人目标的检测精度;使用k-means++算法重新生成数据集锚框,提高网络收敛速度.构建行人检测数据集并分为训练集和测试集,对DOEYT算法的性能进行试验验证.结果表明,非对称最大池化、Hardswish函数、GC自注意力机制分别使平均准确率AP提高14.4%、7.9%、10.8%;DOEYT算法在测试集上检测的平均准确率高达91.2%,检测速度为103帧/s,可见该算法可快速准确地检测行人,降低交通事故发生的风险.
文摘Retinal fundus images are used to discover many diseases.Several Machine learning algorithms are designed to identify the Glaucoma disease.But the accuracy and time consumption performance were not improved.To address this problem Max Pool Convolution Neural Kuan Filtered Tobit Regres-sive Segmentation based Radial Basis Image Classifier(MPCNKFTRS-RBIC)Model is used for detecting the Glaucoma and Stargardt’s disease by early period using higher accuracy and minimal time.In MPCNKFTRS-RBIC Model,the ret-inal fundus image is considered as an input which is preprocessed in hidden layer 1 using weighted adaptive Kuanfilter.Then,preprocessed retinal fundus is given for hidden layer 2 for extracting the features like color,intensity,texture with higher accuracy.After extracting these features,the Tobit Regressive Segmenta-tion process is performed by hidden layer 3 for partitioning preprocessed image within more segments by analyzing the pixel with the extracted features of the fundus image.Then,the segmented image was given to output layer.The radial basis function analyzes the testing image region of a particular class as well as training image region with higher accuracy and minimum time consumption.Simulation is performed with retinal fundus image dataset with various perfor-mance metrics namely peak signal-to-noise ratio,accuracy and time,error rate concerning several retina fundus image and image size.
