YOLSO: You Only Look Small Object

Small object detection is challenging and far from satisfactory. Most general object detectors suffer from two critical issues with small objects: (1) Feature extractor based on classification network cannot express the characteristics of small objects reasonably due to insufficient appearance information of targets and a large amount of background interference around them. (2) The detector requires a much higher location accuracy for small objects than for general objects. This paper proposes an effective and efficient small object detector YOLSO to address the above problems. For feature representation, we analyze the drawbacks in previous backbones and present a Half-Space Shortcut(HSSC) module to build a background-aware backbone. Furthermore, a coarse-to-fine Feature Pyramid Enhancement(FPE) module is introduced for layer-wise aggregation at a granular level to enhance the semantic discriminability. For loss function, we propose an exponential L1 loss to promote the convergence of regression, and a focal IOU loss to focus on prime samples with high classification confidence and high IOU. Both of them significantly improves the location accuracy of small objects. The proposed YOLSO sets state-of-the-art results on two typical small object datasets, MOCOD and VeDAI, at a speed of over 200 FPS. In the meantime, it also outperforms the baseline YOLOv3 by a wide margin on the common COCO dataset.     

Bovidae-based gelatin: Extractions method,physicochemical and functional properties,applications, and future trends

Gelatin is one of the most important multifunctional biopolymers and is widely used as an essential ingredient in food, pharmaceutical, and cosmetics. Porcine gelatin is regarded as the leading source of gelatin globally then followed by bovine gelatin. Porcine sources are favored over other sources since they are less expensive. However, porcine gelatin is religiously prohibited to be consumed by Muslims and the Jewish community. It is predicted that the global demand for gelatin will increase significantly in the future. Therefore, a sustainable source of gelatin with efficient production and free of disease transmission must be developed. The highest quality of Bovidae-based gelatin (BG) was acquired through alkaline pretreatment, which displayed excellent physicochemical and rheological properties. The utilization of mammalian- and plant-based enzyme significantly increased the gelatin yield. The emulsifying and foaming properties of BG also showed good stability when incorporated into food and pharmaceutical products. Manipulation of extraction conditions has enabled the development of custom-made gelatin with desired properties. This review highlighted the various modifications of extraction and processing methods to improve the physicochemical and functional properties of Bovidae-based gelatin. An in-depth analysis of the crucial stage of collagen breakdown is also discussed, which involved acid, alkaline, and enzyme pretreatment, respectively. In addition, the unique characteristics and primary qualities of BG including protein content, amphoteric property, gel strength, emulsifying and viscosity properties, and foaming ability were presented. Finally, the applications and prospects of BG as the preferred gelatin source globally were outlined.     

Enhanced wide-activated residual network for efficient and accurate image deblocking

To save bandwidth and storage space as well as speed up data transmission, people usually perform lossy compression on images. Although the JPEG standard is a simple and effective compression method, it usually introduces various visually unpleasing artifacts, especially the notorious blocking artifacts. In recent years, deep convolutional neural networks (CNNs) have seen remarkable development in compression artifacts reduction. Despite the excellent performance, most deep CNNs suffer from heavy computation due to very deep and wide architectures. In this paper, we propose an enhanced wide-activated residual network (EWARN) for efficient and accurate image deblocking. Specifically, we propose an enhanced wide-activated residual block (EWARB) as basic construction module. Our EWARB gives rise to larger activation width, better use of interdependencies among channels, and more informative and discriminative non-linearity activation features without more parameters than residual block (RB) and wide-activated residual block (WARB). Furthermore, we introduce an overlapping patches extraction and combination (OPEC) strategy into our network in a full convolution way, leading to large receptive field, enforced compatibility among adjacent blocks, and efficient deblocking. Extensive experiments demonstrate that our EWARN outperforms several state-of-the-art methods quantitatively and qualitatively with relatively small model size and less running time, achieving a good trade-off between performance and complexity.     

Feature selection methods for root-cause analysis among top-level product attributes

Manufacturing companies not only strive to deliver flawless products but also monitor product failures in the field to identify potential quality issues. When product failures occur, quality engineers must identify the root cause to improve any affected product and process. This root-cause analysis can be supported by feature selection methods that identify relevant product attributes, such as manufacturing dates with an increased number of product failures. In this paper, we present different methods for feature selection and evaluate their ability to identify relevant product attributes in a root-cause analysis. First, we compile a list of feature selection methods. Then, we summarize the properties of product attributes in warranty case data and discuss these properties regarding the challenges they pose for machine learning algorithms. Next, we simulate datasets of warranty cases, which emulate these product properties. Finally, we compare the feature selection methods based on these simulated datasets. In the end, the univariate filter information gain is determined to be a suitable method for a wide range of applications. The comparison based on simulated data provides a more general result than other publications, which only focus on a single use case. Due to the generic nature of the simulated datasets, the results can be applied to various root-cause analysis processes in different quality management applications and provide a guideline for readers who wish to explore machine learning methods for their analysis of quality data.     

Effect of wind condition on unintended hydrogen release in a hydrogen refueling station

Ambient condition, especially the wind condition, is an important factor to determine the behavior of hydrogen diffusion during hydrogen release. However, only few studies aim at the quantitative study of the hydrogen diffusion in a wind-exist condition. And very little researches aiming at the variable wind condition have been done. In this paper, the hydrogen diffusion in different wind condition which including the constant wind velocity and the variable wind velocity is investigated numerically. When considering the variable wind velocity, the UDF (user defined function) is compiled. Characteristics of the FGC (flammable gas cloud) and the HMF (hydrogen mass fraction) are analyzed in different wind condition and comparisons are made with the no-wind condition. Results indicate that the constant wind velocity and the variable wind velocity have totally different effect for the determination of hydrogen diffusion. Comparisons between the constant wind velocity and the variable wind velocity indicate that the variable wind velocity may cause a more dangerous situation since there has a larger FGC volume. More importantly, the wind condition has a non-negligible effect when considering the HMF along the radial direction. As the wind velocity increases, the distribution of the HMF along the radial direction is not Gaussian anymore when the distance between the release hole and the observation line exceeds to a critical value. This work can be a supplement of the research on the hydrogen release and diffusion and a valuable reference for the researchers.     

A robust model-free controller for a three-phase grid-connected photovoltaic system based on ultra-local model

In this paper, a robust model-free controller for a grid-connected photovoltaic (PV) system is designed. The system consists of a PV generator connected to a three-phase grid by a DC/AC converter. The control objectives of the overall system are to extract maximum power from the PV source, to control reactive power exchange and to improve the quality of the current injected into the grid. The model-free control technique is based on the use of an ultra-local model instead of the dynamic model of the overall system. The local model is continuously updated based on a numerical differentiator using only the input–output behavior of the controlled system. The model-free controller consists of a classical feedback controller and a compensator for the effects of internal parameter changes and external disturbances. Simulation results illustrate the efficiency of the controller for grid-connected PV systems.     

基于面元素重排的三维网格模型数据隐藏算法

目的 为了解决目前三维数据隐藏算法不能兼顾无失真和盲提取的问题，提出一种新的完全无失真的三维网格模型数据隐藏盲算法。方法 首先使用混沌逻辑映射选择嵌入与提取模式，保证数据的安全性。然后利用面元素重排，完全不会造成三维模型失真的性质，通过不同嵌入模式规则对三角面元素进行重排，以嵌入秘密数据。接收端则可根据相应的提取模式规则提取秘密数据。结果 仿真结果与分析表明，该算法不会对三维模型造成任何失真，嵌入容量为每顶点2比特，且能抵抗仿射变换攻击、噪声攻击和平滑攻击等。结论 这种三维数据隐藏盲算法无失真，容量大、安全性高、鲁棒性强，适用于三维载体不容修改的情形，如军事、医学、秘密通信和版权保护等。     

Prediction of Flooding in Packed Liquid-Liquid and High-Pressure Extraction Columns Using a Gaussian Process

Reliable prediction of flooding conditions is needed for sizing and operating packed extraction columns. Due to the complex interplay of physicochemical properties, operational parameters and the packing-specific properties, it is challenging to develop accurate semi-empirical or rigorous models with a high validity range. State of the art models may therefore fail to predict flooding accurately. To overcome this problem, a data-driven model based on Gaussian processes is developed to predict flooding for packed liquid-liquid and high-pressure extraction columns. The optimized Gaussian process for the liquid-liquid extraction column results in an average absolute relative error (AARE) of 15.23 %, whereas the algorithm for the high-pressure extraction column results in an AARE of 13.68 %. Both algorithms can predict flooding curves for different packing geometries and chemical systems precisely.     

WSN低功耗低时延路径式协同计算方法

针对云计算应用于无线传感器网络(Wireless Sensor Network,WSN)时延敏感型业务时存在的高传输时延问题,提出了一种WSN低功耗低时延路径式协同计算方法。该方法基于一种云雾网络架构开展研究,该架构利用汇聚节点组成雾计算层;在数据传输过程中基于雾计算层的计算能力分步骤完成任务计算,降低任务处理时延;由于汇聚节点计算能力较弱,时延降低将导致能耗增加,WSN工作寿命减短,为此提出能耗约束下的任务映射策略,并利用离散二进制粒子群优化(Binary Particle Swarm Optimization,BPSO)算法解决能耗约束下的时延优化问题。仿真结果表明,在相同的能耗约束下,对比其他算法,基于BPSO算法得出的映射方案能有效降低业务处理时延,满足时延敏感型业务的需求。