Dilated and soft attention-guided convolutional neural network for breast cancer histology images classification

Breast cancer is one of the most common types of cancer in women, and histopathological imaging is considered the gold standard for its diagnosis. However, the great complexity of histopathological images and the considerable workload make this work extremely time-consuming, and the results may be affected by the subjectivity of the pathologist. Therefore, the development of an accurate, automated method for analysis of histopathological images is critical to this field. In this article, we propose a deep learning method guided by the attention mechanism for fast and effective classification of haematoxylin and eosin-stained breast biopsy images. First, this method takes advantage of DenseNet and uses the feature map's information. Second, we introduce dilated convolution to produce a larger receptive field. Finally, spatial attention and channel attention are used to guide the extraction of the most useful visual features. With the use of fivefold cross-validation, the best model obtained an accuracy of 96.47% on the BACH2018 dataset. We also evaluated our method on other datasets, and the experimental results demonstrated that our model has reliable performance. This study indicates that our histopathological image classifier with a soft attention-guided deep learning model for breast cancer shows significantly better results than the latest methods. It has great potential as an effective tool for automatic evaluation of digital histopathological microscopic images for computer-aided diagnosis.     

A charge sharing–based switching scheme for SAR ADCs

This paper presents an energy-efficient switching scheme for successive approximation register (SAR) analogue-to-digital converter (ADC). The proposed scheme employs charge recycling method to keep the capacitor arrays free of transitional energy between bit generations except reset phase. In comparison with the conventional switching scheme, the proposed one achieves 100% transitional energy saving without considering reset phase. In addition, configuration of a 10-bit SAR ADC shows that the proposed switching scheme reduces the capacitor area by 25% compared with the conventional switching scheme.     

基于改进粗糙集理论的SAR图像目标增强

图像增强在提高SAR图像舰船目标检测精度方面具有十分重要的意义。由于传统算法不能很好地对SAR图像进行目标增强，提出了基于改进粗糙集理论和引力场强度的目标增强算法。通过借鉴引力场相关理论知识，将粗糙集条件属性集中的梯度属性改进为引力场强度属性，从而实现对原图像的目标增强。与其他算法进行了实验比较，结果表明提出的改进算法相比于其他算法更适用于SAR图像特性，能更好地对舰船目标像素进行针对性增强，具备一定的工程应用价值。     

Structure of presented stimuli influences gazing behavior and choice

Well-structured stimuli presentation is essential in eye-tracking research to test predefined hypotheses reliably and to conduct relevant gazing behavior studies. Several bottom-up factors associated with stimuli presentation (such as stimuli orientation, size etc.) can influence gazing behavior. However, only a small number of scientific papers address these factors in a sensory and consumer science context and thus provide guidance to practitioners. The two presented eye-tracking studies on food images aimed at evaluating the effect of the bottom-up factors stimulus size, background of the picture, orientation of food product presentation, the evaluated products and the number of alternatives. Significant effects of product group were found in the case of all eye-movement parameters except time to first fixation and first fixation duration. In contrary, orientation significantly influenced only the time to first fixation and first fixation duration parameters. Stimulus size significantly increased fixation and dwell count, while background showed no significant effects. Furthermore, significant relationships were found between the number of presented images and eye-movement and decision time. Less time was needed in 2AFC (alternative forced choice test), 3AFC and 4AFC and significantly more time was needed to choose one alternative out of 7AFC and 8AFC. The results of the two studies show that the investigated bottom-up factors can significantly influence gazing behavior, and therefore need to be carefully considered when planning or comparing results of eye-tracking experiments.     

改进HSR-FCN的服装图像识别分类算法研究

目前网络上的服装图像数量增长迅猛，对于大量服装图像实现智能分类的需求日益增加。将基于区域的全卷积网络（Region-Based Fully Convolutional Networks，R-FCN）引入到服装图像识别中，针对服装图像分类中网络训练时间长、形变服装图像识别率低的问题，提出一种新颖的改进框架HSR-FCN。新框架将R-FCN中的区域建议网络和HyperNet网络相融合，改变图片特征学习方式，使得HSR-FCN可以在更短的训练时间内达到更高的准确率。在模型中引入了空间转换网络，对输入服装图像和特征图进行了空间变换及对齐，加强了对多角度服装和形变服装的特征学习。实验结果表明，改进后的HSR-FCN模型有效地加强了对形变服装图像的学习，且在训练时间更短的情况下，比原来的网络模型R-FCN平均准确率提高了大约3个百分点，达到96.69%。     

基于DRGAN和支持向量机的合成孔径雷达图像目标识别

为解决SAR图像目标识别中样本缺乏和方位角敏感问题,提出了一种基于DRGAN和SVM的SAR图像目标识别算法。首先,采用多尺度分形特征对SAR图像进行增强,经过分割得到目标二值图像,基于Hu二阶矩估计目标的方位角。然后对估计得到的目标方位角进行量化编码,结合原始图像作为输入,对设计的DRGAN模型参数进行训练与优化。由于DRGAN中的深度生成模型将目标姿态与外观表示进行解耦设计,故可利用该模型将SAR图像样本变换到同一方位角区间。基于变换后的训练样本分别提取归一化灰度特征,利用SVM训练分类器。采用MSTAR数据集在多个不同操作条件下对提出的算法进行测试,实验结果表明,在带变体的标准操作条件下,能够达到97.97%的分类精度,优于部分基于CNN模型的分类精度,在4种扩展操作条件下的分类精度分别为97.83%,91.77%,97.11%和97.04%,均优于传统方法的分类精度。在SAR图像目标方位角估计存在一定误差的情况下,训练得到的GAN模型作为SAR图像目标旋转估计器,能够使得在不进行复杂样本预处理的前提下,仍然取得较高的SAR图像目标识别精度。