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.     

No-Reference Video Quality Assessment using novel hybrid features and two-stage hybrid regression for score level fusion

This paper presents a novel No-Reference Video Quality Assessment (NR-VQA) model that utilizes proposed 3D steerable wavelet transform-based Natural Video Statistics (NVS) features as well as human perceptual features. Additionally, we proposed a novel two-stage regression scheme that significantly improves the overall performance of quality estimation. In the first stage, transform-based NVS and human perceptual features are separately passed through the proposed hybrid regression scheme: Support Vector Regression (SVR) followed by Polynomial curve fitting. The two visual quality scores predicted from the first stage are then used as features for the similar second stage. This predicts the final quality scores of distorted videos by achieving score level fusion. Extensive experiments were conducted using five authentic and four synthetic distortion databases. Experimental results demonstrate that the proposed method outperforms other published state-of-the-art benchmark methods on synthetic distortion databases and is among the top performers on authentic distortion databases. The source code is available at https://github.com/anishVNIT/two-stage-vqa.     

基于声源定位的高校在线巡课系统设计与研究

随着网络建设以及信息化教学方法在高校教学过程中的应用普及,越来越多的高校使用在线巡课系统对教师的教学过程进行跟踪和管理,以便发现课堂教学中的亮点、问题和不足。文章提出的在线巡课系统,基于声源定位的技术,对传统的在线巡课系统做出了改进,解决了已有巡课系统中“只闻其声,不见其人”的问题,能够更直观地跟踪到教师的教学过程,包括教学行为以及师生互动过程,有效提升教务人员巡课效果和体验感受。     

儿童用品TBT通报数据可视化预警研究

探索采用数据可视化技术分析儿童用品TBT通报数据,以可视化图形图像呈现通报热点并揭示趋势信息,提出对策与建议,助力为儿童用品产业升级、TBT预警数据分析和信息传播工作提供新思路,提高中小企业的国外市场准入机会。     

Deployment of a cloud pipeline for real-time visual inspection using fast streaming high-definition images

We investigate the challenges of building an end-to-end cloud pipeline for real-time intelligent visual inspection system for use in automotive manufacturing. Current methods of visual detection in automotive assembly are highly labor intensive, and thus prone to errors. An automated process is sought that can operate within the real-time constraints of the assembly line and can reduce errors. Components of the cloud pipeline include capture of a large set of high-definition images from a camera setup at the assembly location, transfer and storage of the images as needed, execution of object detection, and notification to a human operator when a fault is detected. The end-to-end execution must complete within a fixed time frame before the next car arrives in the assembly line. In this article, we report the design, development, and experimental evaluation of the tradeoffs of performance, accuracy, and scalability for a cloud system.     

Effects of apparent image velocity and complexity on the dynamic visual field using a high-speed train driving simulator

Train driving is a highly visual task. The visual capabilities of the train driver affects driving safety and driving performance. Understanding the effects of train speed and background image complexity on the visual behavior of the high-speed train driver is essential for optimizing performance and safety. This study investigated the role of the apparent image velocity and complexity on the dynamic visual field of drivers. Participants in a repeated-measures experiment drove a train at nine different speeds in a state-of-the-art high-speed train simulator. Eye movement analysis indicated that the effect of image velocity on the dynamic visual field of high-speed train driver was significant while image complexity had no effect on it. The fixation range was increasingly concentrated on the middle of the track as the speed increased, meanwhile there was a logarithmic decline in fixation range for areas surrounding the track. The extent of the visual search field decreased gradually, both vertically and horizontally, as the speed of train increased, and the rate of decrease was more rapid in the vertical direction. A model is proposed that predicts the extent of this tunnel vision phenomenon as a function of the train speed.Relevance to industryThis finding can be used as a basis for the design of high-speed railway system and as a foundation for improving the operational procedures of high-speed train driver for safety.     

The generalized discrete ‐centroid problem

The ‐centroid problem or leader–follower problem is generalized considering different customer choice rules where a customer may use facilities belonging to different firms, if the difference in travel distance (or time) is small enough. Assuming essential goods, some particular customer choice rules are analyzed. Linear programming formulations for the generalized ‐medianoid and ‐centroid problems are presented and an exact solution approach is applied. Some computational examples are included.     

Restructuring bank networks after mergers and acquisitions: A capacitated delocation model for closing and resizing branches

During restructuring processes, due to mergers and acquisitions, banks frequently face the problem of having redundant branches competing in the same market. In this work, we introduce a new Capacitated Branch Restructuring Model which extends the available literature in delocation models. It considers both closing down and long term operations׳ costs, and addresses the problem of resizing open branches in order to maintain a constant service level. We consider, as well, the presence of competitors and allow for ceding market share whenever the restructuring costs are prohibitively expensive.We test our model in a real life scenario, obtaining a reduction of about 40% of the network size, and annual savings over 45% in operation costs from the second year on. We finally perform a sensitivity analysis on critical parameters. This analysis shows that the final design of the network depends on certain strategic decisions concerning the redundancy of the branches, as well as their proximity to the demand nodes and to the competitor׳s branches. At the same time, this design is quite robust to changes in the parameters associated with the adjustments on service capacity and with the market reaction.     

Model Selection for Unsupervised Learning of Visual Context

This study addresses the problem of choosing the most suitable probabilistic model selection criterion for unsupervised learning of visual context of a dynamic scene using mixture models. A rectified Bayesian Information Criterion (BICr) and a Completed Likelihood Akaike’s Information Criterion (CL-AIC) are formulated to estimate the optimal model order (complexity) for a given visual scene. Both criteria are designed to overcome poor model selection by existing popular criteria when the data sample size varies from small to large and the true mixture distribution kernel functions differ from the assumed ones. Extensive experiments on learning visual context for dynamic scene modelling are carried out to demonstrate the effectiveness of BICr and CL-AIC, compared to that of existing popular model selection criteria including BIC, AIC and Integrated Completed Likelihood (ICL). Our study suggests that for learning visual context using a mixture model, BICr is the most appropriate criterion given sparse data, while CL-AIC should be chosen given moderate or large data sample sizes.