Ozone Depletion Identification in Stratosphere Through Faster Region-Based Convolutional Neural Network

The concept of classification through deep learning is to build a model that skillfully separates closely-related images dataset into different classes because of diminutive but continuous variations that took place in physical systems over time and effect substantially.This study has made ozone depletion identification through classification using Faster Region-Based Convolutional Neural Network(F-RCNN).The main advantage of F-RCNN is to accumulate the bounding boxes on images to differentiate the depleted and non-depleted regions.Furthermore,image classification’s primary goal is to accurately predict each minutely varied case’s targeted classes in the dataset based on ozone saturation.The permanent changes in climate are of serious concern.The leading causes beyond these destructive variations are ozone layer depletion,greenhouse gas release,deforestation,pollution,water resources contamination,and UV radiation.This research focuses on the prediction by identifying the ozone layer depletion because it causes many health issues,e.g.,skin cancer,damage to marine life,crops damage,and impacts on living being’s immune systems.We have tried to classify the ozone images dataset into two major classes,depleted and non-depleted regions,to extract the required persuading features through F-RCNN.Furthermore,CNN has been used for feature extraction in the existing literature,and those extricated diverse RoIs are passed on to the CNN for grouping purposes.It is difficult to manage and differentiate those RoIs after grouping that negatively affects the gathered results.The classification outcomes through F-RCNN approach are proficient and demonstrate that general accuracy lies between 91%to 93%in identifying climate variation through ozone concentration classification,whether the region in the image under consideration is depleted or non-depleted.Our proposed model presented 93%accuracy,and it outperforms the prevailing techniques.

Diagnosis of primary clear cell carcinoma of the liver based on Faster region-based convolutional neural network

Background:Distinguishing between primary clear cell carcinoma of the liver(PCCCL)and common hepatocellular carcinoma(CHCC)through traditional inspection methods before the operation is difficult.This study aimed to establish a Faster region-based convolutional neural network(RCNN)model for the accurate differential diagnosis of PCCCL and CHCC.Methods:In this study,we collected the data of 62 patients with PCCCL and 1079 patients with CHCC in Beijing YouAn Hospital from June 2012 to May 2020.A total of 109 patients with CHCC and 42 patients with PCCCL were randomly divided into the training validation set and the test set in a ratio of 4:1.The Faster RCNN was used for deep learning of patients’data in the training validation set,and established a convolutional neural network model to distinguish PCCCL and CHCC.The accuracy,average precision,and the recall of the model for diagnosing PCCCL and CHCC were used to evaluate the detection performance of the Faster RCNN algorithm.Results:A total of 4392 images of 121 patients(1032 images of 33 patients with PCCCL and 3360 images of 88 patients with CHCC)were uesd in test set for deep learning and establishing the model,and 1072 images of 30 patients(320 images of nine patients with PCCCL and 752 images of 21 patients with CHCC)were used to test the model.The accuracy of the model for accurately diagnosing PCCCL and CHCC was 0.962(95%confidence interval[CI]:0.931-0.992).The average precision of the model for diagnosing PCCCL was 0.908(95%CI:0.823-0.993)and that for diagnosing CHCC was 0.907(95%CI:0.823-0.993).The recall of the model for diagnosing PCCCL was 0.951(95%CI:0.916-0.985)and that for diagnosing CHCC was 0.960(95%CI:0.854-0.962).The time to make a diagnosis using the model took an average of 4 s for each patient.Conclusion:The Faster RCNN model can accurately distinguish PCCCL and CHCC.This model could be important for clinicians to make appropriate treatment plans for patients with PCCCL or CHCC.

基于CA-FasterR-CNN的甲骨文原始拓片单字分割方法

甲骨文拓片经过长时间的埋藏和侵蚀,变得形态复杂,字体模糊,单字之间缺乏明确的分隔,这给甲骨文识别带来了极大的困难。基于此,本文提出了一种基于坐标注意力机制的快速区域卷积神经网络(Coordinate Attention Mechanism-based Faster Region Convolutional Neural Network,CA-Faster R-CNN)模型以实现对甲骨文拓片图像中的单字分割。通过坐标通道注意力机制的引入,模型能够更加关注甲骨文字形特征,从而提升了对甲骨文图像细节的捕捉能力,最后训练结果框线与标准框线基本重合,证明模型分割效果良好。

Establishment and application of an artificial intelligence diagnosis system for pancreatic cancer with a faster region-based convolutional neural network

Background:Early diagnosis and accurate staging are important to improve the cure rate and prognosis for pancreatic cancer.This study was performed to develop an automatic and accurate imaging processing technique system,allowing this system to read computed tomography(CT)images correctly and make diagnosis of pancreatic cancer faster.Methods:The establishment of the artificial intelligence(AI)system for pancreatic cancer diagnosis based on sequential contrastenhanced CT images were composed of two processes:training and verification.During training process,our study used all 4385 CT images from 238 pancreatic cancer patients in the database as the training data set.Additionally,we used VGG16,which was pretrained in ImageNet and contained 13 convolutional layers and three fully connected layers,to initialize the feature extraction network.In the verification experiment,we used sequential clinical CT images from 238 pancreatic cancer patients as our experimental data and input these data into the faster region-based convolution network(Faster R-CNN)model that had completed training.Totally,1699 images from 100 pancreatic cancer patients were included for clinical verification.Results:A total of 338 patients with pancreatic cancer were included in the study.The clinical characteristics(sex,age,tumor location,differentiation grade,and tumor-node-metastasis stage)between the two training and verification groups were insignificant.The mean average precision was 0.7664,indicating a good training ejffect of the Faster R-CNN.Sequential contrastenhanced CT images of 100 pancreatic cancer patients were used for clinical verification.The area under the receiver operating characteristic curve calculated according to the trapezoidal rule was 0.9632.It took approximately 0.2 s for the Faster R-CNN AI to automatically process one CT image,which is much faster than the time required for diagnosis by an imaging specialist.Conclusions:Faster R-CNN AI is an effective and objective method with high accuracy for the diagnosis of pancreatic cancer.

基于张量虚拟机的快速卷积自动性能优化

卷积神经网络作为深度学习的典型代表,是计算机视觉等任务中最常用的神经网络,然而,卷积运算通常占整个卷积神经网络运行时的90%以上,成为卷积神经网络的性能瓶颈。此外,由于当下硬件的复杂性及工作负载的多样性,之前工作中的一些特定优化往往缺乏性能可移植性。对此,作者提出BlazerML,一个基于张量虚拟机(TVM)模板代码自动生成的开源卷积计算库,可为任何输入形状自动生成高性能的卷积实现。BlazerML是基于Winograd算法实现的,因为该算法是快速卷积算法中性能最高的算法。实验结果表明:BlazerML显著优于当下最先进的开源库。在x86 CPU上运行常见的深度学习网络前向推理分别比OnnxRuntime、MNN和TVM社区版本快1.18~2.47倍、1.18~2.27倍和1.01~1.66倍。在ARMCPU上运行常见深度学习网络的单层推理分别比ACL和FastConv快1.26~6.11倍、1.04~4.28倍。

基于改进Fast R-CNN的红外图像行人检测研究

针对红外图像行人检测任务中行人细节信息少,特征提取计算量大以及易受背景影响等问题,提出了一种改进的Fast R-CNN(快速区域卷积神经网络)红外图像行人检测方法。改进主要涉及两个方面:(1)结合红外图像的特点提出了一种自适应ROI提取算法,在不影响检测准确率的前提下,降低了ROI数量,使得网络的计算量减小;(2)提出了一种加权锚点框的定位机制,基于3种不同宽高比锚点框的检测置信度进行坐标加权,获得更准确的定位框。实验结果表明,本文提出的改进方法与传统的Haar+LBP+HOG+SVM算法及Fast R-CNN算法相比,红外图像行人检测的准确率从80.3%和91.2%提高到92.3%,检测速度从68 ms/f和25 ms/f提高到12 ms/f,提高了系统的性能。

基于流量时空特征的fast-flux僵尸网络检测方法

僵尸网络已成为网络空间安全的主要威胁之一,虽然目前可通过逆向工程等技术来对其进行检测,但是使用了诸如fast-flux等隐蔽技术的僵尸网络可以绕过现有的安全检测并继续存活。现有的fast-flux僵尸网络检测方法主要分为主动和被动两种,前者会造成较大的网络负载,后者存在特征值提取繁琐的问题。因此为了有效检测fast-flux僵尸网络并解决传统检测方法中存在的问题,该文结合卷积神经网络和循环神经网络,提出了基于流量时空特征的fast-flux僵尸网络检测方法。结合CTU-13和ISOT公开数据集的实验结果表明,该文所提检测方法和其他方法相比,准确率提升至98.3%,召回率提升至96.7%,精确度提升至97.5%。

一种基于CNN与FFT‑ELM的输电线路故障识别与定位方法

及时、准确地检测输电线路故障类型与位置是提高电力系统可靠性最重要的问题之一,为此提出一种基于卷积神经网络(convolutional neural networks,CNN)与基于快速傅里叶变换(fast Fourier transform,FFT)的极限学习机(extreme learning machine,ELM)分类模型并行的输电线路故障识别及定位方法。首先,以故障电压时序图作为输入,构建CNN;然后,利用FFT将时域故障电压数据分解,提取各频段的电压峰值与相角作为故障特征样本;接着,以提取的故障特征样本集作为输入,构建ELM网络;最后,通过特征融合层将2个神经网络进行融合,输出故障类型和定位结果。实验结果表明,此方法对输电线路故障识别的准确率为99.95%、故障定位误差在500 m以内、平均误差为263.5 m,可靠性优于其他模型。

基于U-net卷积神经网络的电磁场快速计算方法

有限元法(FEM)是物理场分析常用的方法,但庞大的求解自由度导致FEM计算成本很大。针对FEM计算时间长的问题,构建一种基于U-net卷积神经网络的物理场快速计算方法,将样本数据通过栅格化或点云化处理后作为神经网络的输入和标签数据,通过网络训练实现物理场的快速计算并研究该方法在电磁场计算中的应用。结果表明,该方法能准确有效地预测电势、电场强度、磁感应强度等物理量的分布,且预测时间较FEM仿真计算时间大幅缩短。同时,通过合理选择数据集大小,即使在小数据集下也能有较高的预测精度。

基于RISC-Ⅴ的深度可分离卷积神经网络加速器

人工智能时代,RISC-Ⅴ作为一种新兴的开源精简指令集架构,因其低功耗、模块化、开放性和灵活性等优势,使之成为一种能够适应不断发展的深度学习模型和算法的新平台.但是在硬件资源及功耗受限环境下,基础的RISC-Ⅴ处理器架构无法满足卷积神经网络对高性能计算的需求.为了解决这一问题,本文设计了一个基于RISC-Ⅴ的轻量化深度可分离卷积神经网络加速器,旨在弥补RISC-Ⅴ处理器的卷积计算能力的不足.该加速器支持深度可分离卷积中的两个关键算子,即深度卷积和点卷积,并能够通过共享硬件结构提高资源利用效率.深度卷积计算流水线采用了高效的Winograd卷积算法,并使用2×2数据块组合拼接成4×4数据片的方式来减少传输数据冗余.同时,通过拓展RISC-Ⅴ处理器端指令,使得加速器能够实现更灵活的配置和调用.实验结果表明,相较于基础的RISC-Ⅴ处理器,调用加速器后的点卷积和深度卷积计算取得了显著的加速效果,其中点卷积加速了104.40倍,深度卷积加速了123.63倍.与此同时,加速器的性能功耗比达到了8.7GOPS/W.本文的RISC-Ⅴ处理器结合加速器为资源受限环境下卷积神经网络的部署提供了一个高效可行的选择.

一种基于编码单元快速划分的VVC帧内编码方法

相比于高效视频编码(high efficiency video coding,HEVC)标准,新一代编码标准多功能视频编码(versatile video coding,VVC)引入了很多新的技术,其中包括四叉树(quadtree,QT)和多类型树(multi-type tree,MTT)划分,MTT划分由HEVC中的QT划分延伸而来。新划分方法提高了压缩效率,但导致编码时间急剧增加。为了降低编码复杂度,提出了一种结合深度学习方法和MTT方向早期判决的快速帧内编码算法。首先使用轻量级的卷积神经网络(convolutional neural network,CNN)对QT和部分MTT进行预测划分,其余MTT则采用提前预测MTT划分方向的方法作进一步的优化。实验结果表明,所提方法能够大幅降低编码复杂度,相比于原始编码器的编码时间减少了74.3%,且只有3.3%的码率损失,性能优于对比的方法。

FPGA平台上动态硬件重构的Winograd神经网络加速器

为解决卷积神经网络在FPGA平台上进行硬件加速时存在的资源利用率低和资源受限问题,提出了一种基于FPGA动态部分重构技术和Winograd快速卷积的卷积神经网络加速器。该加速器通过运行时硬件重构对FPGA片上资源进行时分复用,采用流水线方式动态地将各个计算流水段配置到FPGA,各个流水段所对应的卷积计算核心使用Winograd算法进行定制优化,以在解决资源受限问题的同时最大程度地提升计算资源利用效率。针对该加速器架构,进一步构建了组合优化模型,用于搜索在特定FPGA硬件平台上部署特定网络模型的最优并行策略,并使用遗传算法进行设计空间求解。基于Xilinx VC709 FPGA平台对VGG-16网络模型进行部署和分析,综合仿真结果表明,所提出的设计方法能够在资源有限的FPGA上自适应地实现大型神经网络模型,加速器整体性能可以达到1078.3 GOPS,较以往加速器的性能和计算资源利用效率可以分别提升2.2倍和3.62倍。

一种面向自动驾驶路况的目标检测算法

为了对常见的行人和车辆进行检测,采用自行标注的数据集,通过基于faster regionbased convolutional neural network(RCNN)框架的算法进行调参与优化.主干网络采用轻量化网络MobileNetv2,在原生锚框的基础上,区域建议网络(RPN)部分增加2个面积尺度,检测部分使用感兴趣区域(ROI)Align结构,减少特征图映射和均分过程中的误差.实验结果表明:使用faster RCNN目标检测网络,可以有效完成行人和车辆的检测任务,整体效果良好.

基于Faster R-CNN算法的船舶识别检测

目前,检测卫星图像中船舶的常用方法如合成孔径雷达(synthetic-aperture radar,SAR)对多目标仍难以达到精确检测,而更快速的区域卷积神经网络(faster region-based convolutional neural network,Faster R-CNN)算法是一种深度学习算法,用于物体检测和分类时,可以实现高精度实时监测。文章应用Faster R-CNN算法对卫星图像中的船舶进行识别和检测,并与传统尺度不变特征转换(scale-invariant feature transform,SIFT)算法、快速区域卷积神经网络(fast region-based convolutional neural network,Fast R-CNN)算法进行对比。研究结果表明,Faster R-CNN算法比传统SIFT算法和Fast R-CNN算法有更好的收敛速度和识别精度,该算法在船舶识别方面具有较大潜力。

Faster R-CNN内窥镜息肉检测

为了提高内窥镜下肠道息肉检测率,提出一种基于Faster R-CNN的息肉检测方法。在数据预处理阶段,利用中值滤波的非线性滤波特性去除图像反光区域,通过数据增强方法扩充样本数据集。在网络结构上,使用残差网络提取多尺度特征送入区域候选网络,得到息肉候选区域;再通过更快的区域神经网络进行训练直至网络收敛,经过微调得到最终检测网络模型。实验结果表明,该方法能够准确检测息肉并标记息肉位置,准确率为96.9%,召回率为95.8%。

Localization and Classification of Rice-grain Images Using Region Proposals-based Convolutional Neural Network

This paper proposes a solution to localization and classification of rice grains in an image.All existing related works rely on conventional based machine learning approaches.However,those techniques do not do well for the problem designed in this paper,due to the high similarities between different types of rice grains.The deep learning based solution is developed in the proposed solution.It contains pre-processing steps of data annotation using the watershed algorithm,auto-alignment using the major axis orientation,and image enhancement using the contrast-limited adaptive histogram equalization(CLAHE)technique.Then,the mask region-based convolutional neural networks(R-CNN)is trained to localize and classify rice grains in an input image.The performance is enhanced by using the transfer learning and the dropout regularization for overfitting prevention.The proposed method is validated using many scenarios of experiments,reported in the forms of mean average precision(mAP)and a confusion matrix.It achieves above 80%mAP for main scenarios in the experiments.It is also shown to perform outstanding,when compared to human experts.

Small objects detection in UAV aerial images based on improved Faster R-CNN

In order to solve the problem of small objects detection in unmanned aerial vehicle(UAV)aerial images with complex background,a general detection method for multi-scale small objects based on Faster region-based convolutional neural network(Faster R-CNN)is proposed.The bird’s nest on the high-voltage tower is taken as the research object.Firstly,we use the improved convolutional neural network ResNet101 to extract object features,and then use multi-scale sliding windows to obtain the object region proposals on the convolution feature maps with different resolutions.Finally,a deconvolution operation is added to further enhance the selected feature map with higher resolution,and then it taken as a feature mapping layer of the region proposals passing to the object detection sub-network.The detection results of the bird’s nest in UAV aerial images show that the proposed method can precisely detect small objects in aerial images.

基于Faster RCNN的行人检测方法

借鉴目标识别领域的快速区域卷积神经网络(Fast RCNN),提出了基于Faster RCNN的行人检测方法,利用CNN提取图像特征,通过聚类和构建区域建议网络(RPN)提取可能含有行人的区域,利用检测网络对目标区域进行判别和分类,并在INRIA数据集中进行了测试验证。实验结果表明:相比基于可变形的组件模型(DPM)的行人检测方法,提出的方法,在测试集上检测准确度达到92. 7%,相比现有的其他方法,其检测效果更好。

Object detection of artifact threaded hole based on Faster R-CNN

In order to improve the accuracy of threaded hole object detection,combining a dual camera vision system with the Hough transform circle detection,we propose an object detection method of artifact threaded hole based on Faster region-ased convolutional neural network(Faster R-CNN).First,a dual camera image acquisition system is established.One industrial camera placed at a high position is responsible for collecting the whole image of the workpiece,and the suspected screw hole position on the workpiece can be preliminarily selected by Hough transform detection algorithm.Then,the other industrial camera is responsible for collecting the local images of the suspected screw holes that have been detected by Hough transform one by one.After that,ResNet50-based Faster R-CNN object detection model is trained on the self-built screw hole data set.Finally,the local image of the threaded hole is input into the trained Faster R-CNN object detection model for further identification and location.The experimental results show that the proposed method can effectively avoid small object detection of threaded holes,and compared with the method that only uses Hough transform or Faster RCNN object detection alone,it has high recognition and positioning accuracy.

基于改进Fast-SCNN的塑瓶气泡缺陷实时分割算法

在医用塑瓶的瓶身气泡检测时,瓶身气泡位置的任意性、气泡大小的不确定性以及气泡特征与瓶身特征之间的相似性增加了气泡缺陷的检测难度。针对上述气泡缺陷检测难点问题,提出了一种基于改进快速分割卷积神经网络(Fast-SCNN)的实时分割算法。该分割算法的基础框架为Fast-SCNN,而为弥补原有网络分割尺寸的鲁棒性不足,借鉴了SENet的通道间信息的利用与多级跳跃连接的思想,具体为网络进一步下采样提取深层特征,在解码阶段将上采样操作融合SELayer模块,同时增加两次与网络浅层的跳跃连接。设计四组对比实验,在气泡数据集上以平均交并比(MIoU)与算法单张分割时间作为评价指标。实验结果表明,改进Fast-SCNN的综合性能最好,其MIoU为97.08%,其预处理后的医用塑瓶的平均检测时间为24.4 ms,其边界分割准确率较Fast-SCNN提升了2.3%,增强了对微小气泡的分割能力,而且该网络的MIoU相较现有的U-Net提升了0.27%,时间上降低了7.5 ms,综合检测性能远超过全卷积神经网络(FCN-8s)。该算法能够有效地对较小的、边缘不清晰的气泡进行分割,满足对气泡缺陷实时分割检测的工程要求。