深度特征筛选及融合的人脸识别方法研究

针对人脸识别问题,提出采用深度特征筛选及融合的方法.采用卷积神经网络(CNN)学习人脸图像的多层次深度特征.对于所有的深度特征矢量,使用斯皮尔曼等级相关系数筛选其中有效部分.基于支持向量机(SVM)对筛选得到的任一深度特征矢量进行分类决策,并基于线性加权融合对它们的结果进行融合,最终确定待识别样本的人脸类别.基于ORL...     

穿墙雷达多维参数人体姿态识别方法

现有的人体姿态识别方案大多数是从单一的角度来考察人体的姿态特征,但是仅采用距离像很难体现人体关节的位置信息,仅提取微多普勒特征有时会覆盖掉径向速度不明显的特征.为此,本文首先利用慢时间-距离像和慢时间-微多普勒谱图构建出人体姿态的三维张量数据集,扩展了人体姿态的特征维度,然后采用改进型瓶颈残差模块构成的神经网络提高了人...     

Machine Learning Approaches for Metalloproteins

Metalloproteins are a family of proteins characterized by metal ion binding, whereby the presence of these ions confers key catalytic and ligand-binding properties. Due to their ubiquity among biological systems, researchers have made immense efforts to predict the structural and functional roles of metalloproteins. Ultimately, having a comprehensive understanding of metalloproteins will lead to tangible applications, such as designing potent inhibitors in drug discovery. Recently, there has been an acceleration in the number of studies applying machine learning to predict metalloprotein properties, primarily driven by the advent of more sophisticated machine learning algorithms. This review covers how machine learning tools have consolidated and expanded our comprehension of various aspects of metalloproteins (structure, function, stability, ligand-binding interactions, and inhibitors). Future avenues of exploration are also discussed.     

深空光通信中的图像信标捕获技术

为实现深空光通信链路建立过程中精确的对准,提出了一种深空光通信系统扩展信标的捕获方案。该方案以可视地球图像作为信标,在航天器上存储一幅信标图像作为参考图像,采用天线扫描的方式在各点对所瞄准的区域成像,利用像素扫描的方式,使参考图像和实际探测图像进行相关,在天线扫描完成以后,找出相关性最大的位置,即可认为在该点捕获到地球图像。在计算两图像相关系数的过程中,由于傅里叶梅林变换幅度谱具有伸缩及旋转不变性,因此利用傅里叶梅林变换即可消除两图像相关系数因为旋转和伸缩的影响。利用蒙特卡罗方法随机产生1000个视场,仿真结果表明,3σ内正确捕获到信标图像的概率为99.6%,表明这是一种可行的方法。     

Network Bending: Expressive Manipulation of Generative Models in Multiple Domains

This paper presents the network bending framework, a new approach for manipulating and interacting with deep generative models. We present a comprehensive set of deterministic transformations that can be inserted as distinct layers into the computational graph of a trained generative neural network and applied during inference. In addition, we present a novel algorithm for analysing the deep generative model and clustering features based on their spatial activation maps. This allows features to be grouped together based on spatial similarity in an unsupervised fashion. This results in the meaningful manipulation of sets of features that correspond to the generation of a broad array of semantically significant features of the generated results. We outline this framework, demonstrating our results on deep generative models for both image and audio domains. We show how it allows for the direct manipulation of semantically meaningful aspects of the generative process as well as allowing for a broad range of expressive outcomes.     

铁电材料激光刻蚀系统中的PID微机控制技术

国内外已研究的网格化技术包括离子束刻蚀和激光化学辅助刻蚀。由于铁电体材料不易减薄至20μm以下,而且像元中心距也在逐步减小,采用离子束刻蚀技术实施网格化很困难。因此,为实现热释电非制冷焦平面探测器的网格化,减小热串音,提高器件的空间分辨率,采用了一套基于PID微机控制技术的激光化学辅助刻蚀系统对铁电材料进行了网格化刻蚀,较好地解决了大列阵铁电材料探测器的网状技术这一难题。     

Application of Deep Eutectic Solvents in the Synthesis of Substituted 2-Mercaptoquinazolin-4(3H)-Ones: A Comparison of Selected Green Chemistry Methods

In this study, deep eutectic solvents (DESs) were used as green and eco-friendly media for the synthesis of substituted 2-mercaptoquinazolin-4(3H)-ones from different anthranilic acids and aliphatic or aromatic isothiocyanates. A model reaction on anthranilic acid and phenyl isothiocyanate was performed in 20 choline chloride-based DESs at 80 °C to find the best solvent. Based on the product yield, choline chloride:urea (1:2) DES was found to be the most effective, while DESs acted both as solvents and catalysts. Desired compounds were prepared with moderate to good yields using stirring, microwave-assisted, and ultrasound-assisted synthesis. Significantly, higher yields were obtained with mixing and ultrasonication (16–76%), while microwave-induced synthesis showed lower effectiveness (13–49%). The specific contribution of this research is the use of DESs in combination with the above-mentioned green techniques for the synthesis of a wide range of derivatives. The structures of the synthesized compounds were confirmed by ¹H and ¹³C NMR spectroscopy.     

Fusion Domain-Adaptation CNN Driven by Images and Vibration Signals for Fault Diagnosis of Gearbox Cross-Working Conditions

The vibration signal of gearboxes contains abundant fault information, which can be used for condition monitoring. However, vibration signal is ineffective for some non-structural failures. In order to resolve this dilemma, infrared thermal images are introduced to combine with vibration signals via fusion domain-adaptation convolutional neural network (FDACNN), which can diagnose both structural and non-structural failures under various working conditions. First, the measured raw signals are converted into frequency and squared envelope spectrum to characterize the health states of the gearbox. Second, the sequences of the frequency and squared envelope spectrum are arranged into two-dimensional format, which are combined with infrared thermal images to form fusion data. Finally, the adversarial network is introduced to realize the state recognition of structural and non-structural faults in the unlabeled target domain. An experiment of gearbox test rigs was used for effectiveness validation by measuring both vibration and infrared thermal images. The results suggest that the proposed FDACNN method performs best in cross-domain fault diagnosis of gearboxes via multi-source heterogeneous data compared with the other four methods.     

A Probabilistic Re-Intepretation of Confidence Scores in Multi-Exit Models

In this paper, we propose a new approach to train a deep neural network with multiple intermediate auxiliary classifiers, branching from it. These ‘multi-exits’ models can be used to reduce the inference time by performing early exit on the intermediate branches, if the confidence of the prediction is higher than a threshold. They rely on the assumption that not all the samples require the same amount of processing to yield a good prediction. In this paper, we propose a way to train jointly all the branches of a multi-exit model without hyper-parameters, by weighting the predictions from each branch with a trained confidence score. Each confidence score is an approximation of the real one produced by the branch, and it is calculated and regularized while training the rest of the model. We evaluate our proposal on a set of image classification benchmarks, using different neural models and early-exit stopping criteria.