Multi-label,multi-task CNN approach for context-based emotion recognition

This paper proposes a new deep learning architecture for context-based multi-label multi-task emotion recognition. The architecture is built from three main modules: (1) a body features extraction module, which is a pre-trained Xception network, (2) a scene features extraction module, based on a modified VGG16 network, and (3) a fusion-decision module. Moreover, three categorical and three continuous loss functions are compared in order to point out the importance of the synergy between loss functions when it comes to multi-task learning. Then, we propose a new loss function, the multi-label focal loss (MFL), based on the focal loss to deal with imbalanced data. Experimental results on EMOTIC dataset show that MFL with the Huber loss gave better results than any other combination and outperformed the current state of art on the less frequent labels.     

RXDNFuse: A aggregated residual dense network for infrared and visible image fusion

This study proposes a novel unsupervised network for IR/VIS fusion task, termed as RXDNFuse, which is based on the aggregated residual dense network. In contrast to conventional fusion networks, RXDNFuse is designed as an end-to-end model that combines the structural advantages of ResNeXt and DenseNet. Hence, it overcomes the limitations of the manual and complicated design of activity-level measurement and fusion rules. Our method establishes the image fusion problem into the structure and intensity proportional maintenance problem of the IR/VIS images. Using comprehensive feature extraction and combination, RXDNFuse automatically estimates the information preservation degrees of corresponding source images, and extracts hierarchical features to achieve effective fusion. Moreover, we design two loss function strategies to optimize the similarity constraint and the network parameter training, thus further improving the quality of detailed information. We also generalize RXDNFuse to fuse images with different resolutions and RGB scale images. Extensive qualitative and quantitative evaluations reveal that our results can effectively preserve the abundant textural details and the highlighted thermal radiation information. In particular, our results form a comprehensive representation of scene information, which is more in line with the human visual perception system.     

改进RetinaFace的自然场景口罩佩戴检测算法

新型冠状病毒可以通过空气中的飞沫、气溶胶等载体进行传播，在公共场所下正确佩戴口罩可以有效地防止病毒的传播。提出了一种自然场景下人脸口罩佩戴检测方法，对RetinaFace算法进行了改进，增加了人脸口罩佩戴检测任务，优化了损失函数。在特征金字塔网络中引入了一种改进的自注意力机制，增强了特征图的表达能力。建立了包含3 000张图片的数据集，并进行手工标注，用于网络训练。实验结果表明该算法可以有效进行口罩佩戴检测，在自然场景视频中也取得了不错的检测效果。     

结合注意力的3D卷积网络脑胶质瘤分割算法

为了提升脑胶质瘤分割精度，提出一种结合注意力机制的3D卷积神经网络算法。输入3个不同尺度的图像块，经过9个卷积层和1个分类层后得到3个不同的分类结果，将分类结果与注意力学习到的权重相乘并逐体素相加得到输出。此外该算法采用了一种混合Dice损失函数与Focal损失函数的超参数损失函数。实验表明，该算法的Dice系数在整体区域、核心区域以及增强区域分别达到了95.31%、80.12%、82.25%。与已有的一种脑胶质瘤分割算法deepmedic相比，整体区域、核心区域以及增强区域的Dice系数分别提升了3%、2%、6%。在脑胶质瘤分割方面，具有重要的临床意义。     

基于传统清管球接收器的滤杂防堵工艺设计

清管球回收操作过程中,清管球接收器的泄压、排液和惰化导致泥、砂、垢等杂质积聚在清管球接收器中,并随油气物流进入闭排系统管线,造成杂质流失、闭排系统管线堵塞和压力无法完全释放,影响对海底油气管道腐蚀、析蜡、结垢及地层出砂等情况的判断,存在很大的安全隐患。通过对清管球接收器实施工艺改造,使清管球接收器具备滤杂防堵功能和自动取样功能,有效解决了闭排管线堵塞、杂质流失和安全隐患问题。该工艺研究对于海上油气田海底管线的运营和维护具有较高的借鉴意义。     

基于艾默生DeltaV R6新功能块的设备控制模块开发

利用艾默生控制系统旧版本软件,开发开关量设备控制模块比较繁琐,需要借助大量的逻辑功能块与条件判断功能块进行辅助开发,程序结构复杂,逻辑不清晰,容易出错。新版本软件推出的新功能块解决了以上问题。以某火法铜冶炼工程为例,结合冶炼工艺特点,程序开发过程中通过解决不熟悉新功能块特点应用难问题,风机、泵、电动阀等设备控制模块开发难问题,开发出了满足控制要求的各种设备控制模块。利用新版本软件的新功能块,开发的设备控制模块在火法铜冶炼得到成功应用。     

High-speed,two-dimensional digital image correlation algorithm using heterogeneous (CPU-GPU) framework

Two-dimensional digital image correlation (2D-DIC) is an experimental technique used to measure in-plane displacement of a test specimen. Real-time measurement of full-field displacement data is challenging due to enormous computational load of the algorithm. In order to improve the computational speed, the focus of recent research works has been on the approach of parallelization across subsets within image pairs using graphics processing unit (GPU). But alternate GPU-based parallelization approaches to improve the performance of this algorithm as per the order of data processing have not been explored. To address this research gap, our method utilizes parallelism within a subset as well as across subsets for each computation step in an iteration cycle. A heterogeneous (CPU-GPU) framework in combination with a pyramid-based initial values estimation for subsets (in parallel) is proposed in this work. The precompute steps of the proposed framework are implemented using CPU, whereas the main iterative steps are realized using GPU. It is demonstrated that the overall computational speed of the proposed heterogeneous framework improves by compared to a sequential CPU-based implementation for a pair of gray-scale images with a resolution of pixels. As an important milestone, feasibility to measure deformations in real time ( 1 s) is manifested in this study.     

Analyzing dissipation effects on the non-classicality witnessing by a qubit via the entangled state representation

ABSTRACT

In this paper, we study detection of the state non-classicality for a quantum harmonic oscillator by a qubit in the presence of dissipation effects. We show that dissipation can enhance the effectiveness of the method in case of using the corrected form of the related nonclassicality witness. Such an improvement is attributed to the fact that dissipation leads to probing a surface, instead of a curve, of the complex plane for non-classicality condition on normally-ordered characteristic function.     

Comparative study on pyrolysis characteristics and kinetics of oleaginous yeast and algae

The pyrolysis processes of oleaginous yeast and algae were studied and compared using a non-isothermal thermogravimetric analyzer at heating rates of 10–50 °C/min, and the most probable mechanism function and kinetic analyses of the main stage of pyrolysis were carried out by the Popuse method, Starink method, and Fridemen method. The main pyrolysis stage of the samples could be described by the Jander equation and Z–L–T equation and the activation energy of the three biomass was 108–117, 107–121 and 93–108 kJ/mol, respectively. For the three kinds of biomass, the DTG curves were divided based on the four pseudo-components by performing Gaussian fitting which are carbohydrates, proteins, lipids, others, and the weight coefficients of them could be identified. The activation energy of each pseudo-component was obtained in the range of 58.36–140.44 kJ/mol by the Kissinger method. The four-pseudo-component model based on Gaussian fitting provides effective data for the design of oleaginous yeast and algae thermal decomposition systems and the kinetic analysis of the pyrolysis process.