Geometry Flow-Based Deep Riemannian Metric Learning

Deep metric learning(DML)has achieved great results on visual understanding tasks by seamlessly integrating conventional metric learning with deep neural networks.Existing deep metric learning methods focus on designing pair-based distance loss to decrease intra-class distance while increasing interclass distance.However,these methods fail to preserve the geometric structure of data in the embedding space,which leads to the spatial structure shift across mini-batches and may slow down the convergence of embedding learning.To alleviate these issues,by assuming that the input data is embedded in a lower-dimensional sub-manifold,we propose a novel deep Riemannian metric learning(DRML)framework that exploits the non-Euclidean geometric structural information.Considering that the curvature information of data measures how much the Riemannian(nonEuclidean)metric deviates from the Euclidean metric,we leverage geometry flow,which is called a geometric evolution equation,to characterize the relation between the Riemannian metric and its curvature.Our DRML not only regularizes the local neighborhoods connection of the embeddings at the hidden layer but also adapts the embeddings to preserve the geometric structure of the data.On several benchmark datasets,the proposed DRML outperforms all existing methods and these results demonstrate its effectiveness.

Deep Learning Aided SCL Decoding of Polar Codes with Shifted-Pruning

Recently,a generalized successive cancellation list(SCL)decoder implemented with shiftedpruning(SP)scheme,namely the SCL-SP-ωdecoder,is presented for polar codes,which is able to shift the pruning window at mostωtimes during each SCL re-decoding attempt to prevent the correct path from being eliminated.The candidate positions for applying the SP scheme are selected by a shifting metric based on the probability that the elimination occurs.However,the number of exponential/logarithm operations involved in the SCL-SP-ωdecoder grows linearly with the number of information bits and list size,which leads to high computational complexity.In this paper,we present a detailed analysis of the SCL-SP-ωdecoder in terms of the decoding performance and complexity,which unveils that the choice of the shifting metric is essential for improving the decoding performance and reducing the re-decoding attempts simultaneously.Then,we introduce a simplified metric derived from the path metric(PM)domain,and a custom-tailored deep learning(DL)network is further designed to enhance the efficiency of the proposed simplified metric.The proposed metrics are both free of transcendental functions and hence,are more hardware-friendly than the existing metrics.Simulation results show that the proposed DL-aided metric provides the best error correction performance as comparison with the state of the art.

Hybrid Deep Learning-Improved BAT Optimization Algorithm for Soil Classification Using Hyperspectral Features

Now a days,Remote Sensing(RS)techniques are used for earth observation and for detection of soil types with high accuracy and better reliability.This technique provides perspective view of spatial resolution and aids in instantaneous measurement of soil’s minerals and its characteristics.There are a few challenges that is present in soil classification using image enhancement such as,locating and plotting soil boundaries,slopes,hazardous areas,drainage condition,land use,vegetation etc.There are some traditional approaches which involves few drawbacks such as,manual involvement which results in inaccuracy due to human interference,time consuming,inconsistent prediction etc.To overcome these draw backs and to improve the predictive analysis of soil characteristics,we propose a Hybrid Deep Learning improved BAT optimization algorithm(HDIB)for soil classification using remote sensing hyperspectral features.In HDIB,we propose a spontaneous BAT optimization algorithm for feature extraction of both spectral-spatial features by choosing pure pixels from the Hyper Spectral(HS)image.Spectral-spatial vector as training illustrations is attained by merging spatial and spectral vector by means of priority stacking methodology.Then,a recurring Deep Learning(DL)Neural Network(NN)is used for classifying the HS images,considering the datasets of Pavia University,Salinas and Tamil Nadu Hill Scene,which in turn improves the reliability of classification.Finally,the performance of the proposed HDIB based soil classifier is compared and analyzed with existing methodologies like Single Layer Perceptron(SLP),Convolutional Neural Networks(CNN)and Deep Metric Learning(DML)and it shows an improved classification accuracy of 99.87%,98.34%and 99.9%for Tamil Nadu Hills dataset,Pavia University and Salinas scene datasets respectively.

Weed Classification Using Particle Swarm Optimization and Deep Learning Models

Weed is a plant that grows along with nearly allfield crops,including rice,wheat,cotton,millets and sugar cane,affecting crop yield and quality.Classification and accurate identification of all types of weeds is a challenging task for farmers in earlier stage of crop growth because of similarity.To address this issue,an efficient weed classification model is proposed with the Deep Convolutional Neural Network(CNN)that implements automatic feature extraction and performs complex feature learning for image classification.Throughout this work,weed images were trained using the proposed CNN model with evolutionary computing approach to classify the weeds based on the two publicly available weed datasets.The Tamil Nadu Agricultural University(TNAU)dataset used as afirst dataset that consists of 40 classes of weed images and the other dataset is from Indian Council of Agriculture Research–Directorate of Weed Research(ICAR-DWR)which contains 50 classes of weed images.An effective Particle Swarm Optimization(PSO)technique is applied in the proposed CNN to automa-tically evolve and improve its classification accuracy.The proposed model was evaluated and compared with pre-trained transfer learning models such as GoogLeNet,AlexNet,Residual neural Network(ResNet)and Visual Geometry Group Network(VGGNet)for weed classification.This work shows that the performance of the PSO assisted proposed CNN model is significantly improved the success rate by 98.58%for TNAU and 97.79%for ICAR-DWR weed datasets.

Mesh representation matters:investigating the influence of different mesh features on perceptual and spatial fidelity of deep 3D morphable models

Background Deep 3D morphable models(deep 3DMMs)play an essential role in computer vision.They are used in facial synthesis,compression,reconstruction and animation,avatar creation,virtual try-on,facial recognition systems and medical imaging.These applications require high spatial and perceptual quality of synthesised meshes.Despite their significance,these models have not been compared with different mesh representations and evaluated jointly with point-wise distance and perceptual metrics.Methods We compare the influence of different mesh representation features to various deep 3DMMs on spatial and perceptual fidelity of the reconstructed meshes.This paper proves the hypothesis that building deep 3DMMs from meshes represented with global representations leads to lower spatial reconstruction error measured with L_(1) and L_(2) norm metrics and underperforms on perceptual metrics.In contrast,using differential mesh representations which describe differential surface properties yields lower perceptual FMPD and DAME and higher spatial fidelity error.The influence of mesh feature normalisation and standardisation is also compared and analysed from perceptual and spatial fidelity perspectives.Results The results presented in this paper provide guidance in selecting mesh representations to build deep 3DMMs accordingly to spatial and perceptual quality objectives and propose combinations of mesh representations and deep 3DMMs which improve either perceptual or spatial fidelity of existing methods.

基于CenterNet搭配优化DeepSORT算法的断奶仔猪目标跟踪方法研究

[目的]断奶仔猪是生猪养殖业中的重点关注对象,群养环境下的仔猪精准跟踪是分析仔猪个体行为、监测仔猪个体健康的基础。本文提出1种基于深度学习的断奶仔猪目标跟踪方法。[方法]利用基于中心点的CenterNet检测算法设计断奶仔猪目标检测模型,对DeepSORT算法的检测部分进行优化;训练优化跟踪过程中提取仔猪目标外观特征的模型,结合卡尔曼滤波和匈牙利匹配算法,改进重识别环节,实现具有深度关联度量的断奶仔猪目标实时检测与跟踪。[结果]算法测试结果表明,断奶仔猪目标检测模型的平均精度均值和召回率分别为99.0%和78.6%,多目标跟踪精度MOTA和MOTP指标分别为96.8%和81.8%。[结论]本文所提方法改善了因断奶仔猪外表高度相似性以及黏连遮挡情况导致跟踪困难的问题,可在群养环境中精准跟踪断奶仔猪个体,算法为后续仔猪个体行为分析研究提供技术支撑。

An Efficient Medical Image Deep Fusion Model Based on Convolutional Neural Networks

Medical image fusion is considered the best method for obtaining one image with rich details for efficient medical diagnosis and therapy.Deep learning provides a high performance for several medical image analysis applications.This paper proposes a deep learning model for the medical image fusion process.This model depends on Convolutional Neural Network(CNN).The basic idea of the proposed model is to extract features from both CT and MR images.Then,an additional process is executed on the extracted features.After that,the fused feature map is reconstructed to obtain the resulting fused image.Finally,the quality of the resulting fused image is enhanced by various enhancement techniques such as Histogram Matching(HM),Histogram Equalization(HE),fuzzy technique,fuzzy type,and Contrast Limited Histogram Equalization(CLAHE).The performance of the proposed fusion-based CNN model is measured by various metrics of the fusion and enhancement quality.Different realistic datasets of different modalities and diseases are tested and implemented.Also,real datasets are tested in the simulation analysis.

Kernel Granulometric Texture Analysis and Light RES-ASPP-UNET Classification for Covid-19 Detection

This research article proposes an automatic frame work for detectingCOVID -19 at the early stage using chest X-ray image. It is an undeniable factthat coronovirus is a serious disease but the early detection of the virus presentin human bodies can save lives. In recent times, there are so many research solutions that have been presented for early detection, but there is still a lack in needof right and even rich technology for its early detection. The proposed deeplearning model analysis the pixels of every image and adjudges the presence ofvirus. The classifier is designed in such a way so that, it automatically detectsthe virus present in lungs using chest image. This approach uses an imagetexture analysis technique called granulometric mathematical model. Selectedfeatures are heuristically processed for optimization using novel multi scaling deep learning called light weight residual–atrous spatial pyramid pooling(LightRES-ASPP-Unet) Unet model. The proposed deep LightRES-ASPPUnet technique has a higher level of contracting solution by extracting majorlevel of image features. Moreover, the corona virus has been detected usinghigh resolution output. In the framework, atrous spatial pyramid pooling(ASPP) method is employed at its bottom level for incorporating the deepmulti scale features in to the discriminative mode. The architectural workingstarts from the selecting the features from the image using granulometricmathematical model and the selected features are optimized using LightRESASPP-Unet. ASPP in the analysis of images has performed better than theexisting Unet model. The proposed algorithm has achieved 99.6% of accuracyin detecting the virus at its early stage.

文本分类算法及其应用场景研究综述

随着大数据时代的到来,互联网中的文本信息迎来了井喷式的增长.文本分类作为自然语言处理中最重要的技术之一,其广泛应用于多个领域,如情感分析、新闻分类、自然语言推理、主题标记、抽取式问答、虚假内容检测等.从传统机器学习分类方法理论的深入到深度学习分类方法探索的兴起,相关研究模型与思路也在不断演变,各类新的方法、数据集和评价指标层出不穷,丰富了文本分类领域的研究,取得了卓越的理论成就和应用效果.尽管如此,新技术不断发展和业务应用场景不断丰富,同时,也为文本分类研究带来了许多新的问题与挑战,如数据约束场景中不均衡数据的文本表征学习、小样本场景下的文本分类等.针对当前研究难题与挑战,本文对文本分类方法进行了系统性调研,并对当前方法在实际应用场景中面临的技术挑战和未来的研究方向进行了综合探讨.具体而言,本文主要综述了七部分内容,分别是:(1)对文本分类技术的相关基础知识进行了全面介绍,包括文本分类的常见符号定义、计算范式和文本预处理技术;(2)对基于传统机器学习的文本分类方法进行了详细总结;同时,为了方便读者针对不同的应用场景选择合适的分类模型,本文对不同分类器擅长处理的文本分类难题及方法优劣进行了总结;(3)对基于新兴深度学习的文本分类方法进行了周详梳理,根据领域内代表性技术的核心思想进行分类,在此基础上对不同类别下的主要方法进行描述,同时对其技术的优劣进行了总结;(4)为了方便读者对文本分类模型的有效性进行验证,针对文本分类技术应用最为广泛的七大场景,本文对相关数据集进行了系统性的总结;(5)本文对不同任务目标下的常用的模型评价方法进行详尽介绍,以便对模型性能进行合理的定量评估;(6)基于上述内容,本文对典型应用场景中不同种类文本分类算法进行了性能总结对比;(7)本文分别从数据约束与模型计算两个层面对当前文本分类技术所面临的挑战和未来的重要研究方向进行了总结.本文通过梳理文本分类研究发展脉络,对涉及的代表性技术进行了详细总结和对比分析,有效填补了文本分类领域前沿技术的应用综述空白.

Pseudo-label based semi-supervised learning in the distributed machine learning framework

With the emergence of various intelligent applications,machine learning technologies face lots of challenges including large-scale models,application oriented real-time dataset and limited capabilities of nodes in practice.Therefore,distributed machine learning(DML) and semi-supervised learning methods which help solve these problems have been addressed in both academia and industry.In this paper,the semi-supervised learning method and the data parallelism DML framework are combined.The pseudo-label based local loss function for each distributed node is studied,and the stochastic gradient descent(SGD) based distributed parameter update principle is derived.A demo that implements the pseudo-label based semi-supervised learning in the DML framework is conducted,and the CIFAR-10 dataset for target classification is used to evaluate the performance.Experimental results confirm the convergence and the accuracy of the model using the pseudo-label based semi-supervised learning in the DML framework.Given the proportion of the pseudo-label dataset is 20%,the accuracy of the model is over 90% when the value of local parameter update steps between two global aggregations is less than 5.Besides,fixing the global aggregations interval to 3,the model converges with acceptable performance degradation when the proportion of the pseudo-label dataset varies from 20% to 80%.

基于度量学习的多模态谣言检测

目前主流的多模态谣言检测模型,主要侧重于建模过程中模态的特征提取与拼接方法研究,而各模态局部特征关系、模态内与模态间的信息交互往往被忽略,这在一定程度上影响到了谣言检测的效果。针对该问题,提出了一种基于度量学习的多模态谣言检测方法。考虑到各模态局部特征关系对模态整体特征表示的影响,采用了句法分析和注意力机制技术分别挖掘文本和图片的局部特征关系;同时,将度量学习应用到谣言检测中,通过三元组学习和对比学习找出模态内与模态间的关联信息。在Twitter和Weibo两个公开的数据集上进行了性能测试实验,准确率分别达到92.8%和85.2%,结果表明将各模态局部特征关系、模态内与模态间的信息交互加入谣言检测模型中能够进一步提升谣言检测的精准度。

基于深度学习的视频修复方法综述

视频作为常见的媒体信息之一,目前已在各个领域得到广泛应用。尤其是以抖音等为代表的短视频软件的兴起,使得与视频相关的技术不断迭代更新。其中,视频修复技术是视频处理研究中的一个热点。视频修复技术是利用视频帧内的像素信息和帧间的时域参考信息对视频帧受损的区域进行内容推理并修复,在补全缺损视频、物体移除及视频伪造检测等场景中具有广泛应用前景。该技术可追溯到二十世纪末的老电影修复技术,该任务通常由专业的技术团队逐帧修复完成。而随着数字技术的发展,近年来已有一些人工智能技术用于视频修复,让老电影重获新生。目前,视频修复技术可分为传统方法和基于深度学习的两类方法。其中,传统方法由于缺少对高层语义信息的理解,在场景复杂、缺失区域较大的情况下修复效果不佳;而基于深度学习的方法随着算法框架的优化和图形处理器性能的提升展现了出色的效果,对修复结果的语义结构准确性和时间一致性都有明显的提升。本文在简要回顾传统视频修复方法的基础上,重点分析四类基于深度学习视频修复方法的网络结构、参数模型、性能表现与优缺点;介绍视频修复领域中常用的数据集和评价指标;最后,对视频修复领域现存的问题进行总结并展望未来可能的研究方向。

基于度量学习的电路焊点缺陷检测方法

针对目前电路焊点缺陷检测方法效率低、准确度差、焊点图像样本量小的问题,提出了一种基于度量学习的快速识别焊点缺陷的方法.首先利用工业相机搭配远心镜头获取焊点图像.通过挖掘焊点图像特征,设计交点检测法来分割焊接单元图像,制作焊点缺陷数据集.在此基础上,设计焊点图像全局特征与局部表征提取方法来对焊点的两类特征进行融合,并对注意力机制进行改进,加入到全局特征提取模块中.对焊点缺陷的检测实验结果表明该方法最终实现了准确率达到98.4%,满足焊点缺陷检测的实际生产要求.

基于跨模态近邻损失的可视-红外行人重识别

可视-红外跨模态行人重识别任务的目标是给定一个模态的特定人员图像,在其他不同模态摄像机所拍摄的图像集中进行检索,找出相同人员对应的图像。由于成像方式不同,不同模态的图像之间存在明显的模态差异。为此,从度量学习的角度出发,对损失函数进行改进以获取具有更加辨别性的信息。对图像特征内聚性进行理论分析,并在此基础上提出一种基于内聚性分析和跨模态近邻损失函数的重识别方法,以加强不同模态样本的内聚性。将跨模态困难样本的相似性度量问题转化为跨模态最近邻样本对和同模态样本对的相似性度量,使得网络对模态内聚性的优化更加高效和稳定。对所提方法在全局特征表示的基线网络和部分特征表示的基线网络上进行实验验证结果表明:所提方法对可视-红外行人重识别的预测结果相较于基线方法,平均准确度最高可提升8.44%,证明了方法在不同网络架构中的通用性;同时,以较小的模型复杂度和较低的计算量为代价,实现了可靠的跨模态行人重识别结果。

小样本下基于改进度量学习的轨面状态识别

为解决小样本下轨面状态识别过程中存在的关键特征信息提取不充分、区分度信息易丢失的问题,提出一种基于改进度量学习的轨面状态识别方法。该方法在特征提取网络部分引入金字塔拆分注意力机制,实现特征图空间信息多尺度提取、跨维度通道注意力与空间注意力特征交互,以解决轨面状态样本少导致的关键特征信息提取不充分的问题。利用深度局部拼接符对查询集与各类支撑集特征图进行局部特征两两拼接,代替传统度量学习的全局特征拼接,筛选背景等干扰信息,较大程度地保留有显著区分度的特征信息。在自建小样本轨面状态数据集上进行性能验证,并与常规小样本学习方法进行对比实验,实验结果表明,本文方法能够有效识别轨面状态,识别准确率、精度、召回率、F1值分别达到97.96%、98.61%、98.07%、98.34%,相比于性能较好的小样本学习方法 DN4网络,各项指标分别提升了5.75个百分点、5.83个百分点、5.95个百分点、5.89个百分点。

基于数据特征相关性和自适应差分隐私的深度学习方法研究

基于差分隐私的深度学习隐私保护方法中,训练周期的长度以及隐私预算的分配方式直接制约着深度学习模型的效用.针对现有深度学习结合差分隐私的方法中模型训练周期有限、隐私预算分配不合理导致模型安全性与可用性差的问题,提出一种基于数据特征相关性和自适应差分隐私的深度学习方法(deep learning methods based on data feature Relevance and Adaptive Differential Privacy,RADP).首先,该方法利用逐层相关性传播算法在预训练模型上计算出原始数据集上每个特征的平均相关性;然后,使用基于信息熵的方法计算每个特征平均相关性的隐私度量,根据隐私度量对特征平均相关性自适应地添加拉普拉斯噪声;在此基础上,根据加噪保护后的每个特征平均相关性,合理分配隐私预算,自适应地对特征添加拉普拉斯噪声;最后,理论分析该方法(RADP)满足ε-差分隐私,并且兼顾安全性与可用性.同时,在三个真实数据集(MNIST,Fashion-MNIST,CIFAR-10)上的实验结果表明,RADP方法的准确率以及平均损失均优于AdLM(Adaptive Laplace Mechanism)方法、DPSGD(Differential Privacy with Stochastic Gradient Descent)方法和DPDLIGDO(Differentially Private Deep Learning with Iterative Gradient Descent Optimization)方法,并且RADP方法的稳定性仍能保持良好.

面向电力生产精细化风险解译的高度相似防护工具智能检测技术研究

电力生产通常面临高低电压、强弱电流等复杂工作环境转换,不同作业场景有严格的防护工具使用标准,因此,研究生产作业过程防护工具的精细辨识对保障人员安全及电网安全意义重大。已有研究可实现安全帽、工作服等基础着装类检测,而实际生产中存在大量形态高度相似的实体防护工具,如绝缘手套与线手套、绝缘杆与验电杆等。为此,该文提出一种基于深度代表性度量学习的相似防护工具智能检测方法。将目标类别特征学习转换为以差异化表达不同目标特征距离为目的的嵌入式空间特征学习,得到表征不同目标的深度代表性特征向量,通过计算未知目标与代表性特征向量的距离进行类别判断,最后以现场图像进行试验验证。试验结果表明:所提方法实现了对形态相似防护工具的特征差异表达和精准辨识,相比于常见目标检测模型具有更优越的辨识性能,从而提高电力生产安全风险辨识的精细化水平。

基于深度度量学习的强泛化开关仪表识别算法

针对电厂开关检测方法无法应对现实开集环境,对稀有类别识别准确率低的现状,将目标识别问题转化为相似性度量问题,并提出新算法。新算法基于深度度量学习的三元组网络,利用加入SE Block的ResNet-18提取特征,并利用跨批次挖掘增强学习效果。为评估算法性能,创建了一个包含3300张开关图片的数据集,并使用新算法在该数据集上进行了闭集测试、开集测试、小样本测试。结果表明:新算法在闭集状态下具有良好的区分能力,不仅能准确识别训练集中的类别,还能有效区分训练时未遇到的及出现频率较低的状态。由此表明,该算法不仅适用于现实世界的开集环境,而且能显著提升对小样本数据的识别精度。

深度度量学习综述

随着深度神经网络的兴起,深度度量学习(DML)引起广泛的关注。为了深入了解深度度量学习,首先,整理和分析传统度量学习方法的局限性。其次,从3个类型探讨DML,包括基于样本对、代理和分类的类型:基于样本对的类型包括散度方法、排序方法和基于生成对抗网络(GAN)的方法;基于代理的类型主要从代理样本、类别方面进行讨论;基于分类的类型中主要讨论了跨模态度量学习、类内类间边距问题、超图分类,以及与其他方法(如基于强化学习和基于对抗学习的方法)的结合。再次,介绍评估DML性能的各种指标,同时总结和对比DML在不同任务(包括人脸识别、图像检索和行人重识别等)中的应用。最后,探讨DML面临的挑战,并提出一些可能的解决策略。

基于多域信息融合的深度学习轴承故障诊断方法

针对单一振动信号包含故障信息易被隐藏以及单一深度学习模型诊断能力不强导致轴承故障诊断精度低的问题,提出了一种基于多域信息融合的深度学习故障诊断方法。利用变分模态分解方法(variational mode decomposition,VMD)将原始振动信号分解为多个本征模态函数(intrinsic mode function,IMF)分量,同时对每个IMF分量进行快速傅里叶变换(fast Fourier transformation,FFT)转化为频域样本;然后将多个IMF分量和其对应频域样本分别输入至多个深度度量学习(deep metric learning,DML)模型和深度置信网络(deep belief network,DBN)模型分别进行初步诊断分析,并利用简单软投票法对这些初步诊断结果进行融合从而获取最终诊断结果。最后通过对不同轴承故障的诊断试验分析,结果表明,该研究提出的方法不仅具有较好的诊断效果,而且诊断性能分别优于基于时域和基于频域的信息融合诊断方法。