Medical image classification and diagnosis is currently a hot topic in the field of deep learning. The ACM International Conference on Multimedia and Simula co-hosted the MutilMedia Grand Challenge, which aims to use artificial intelligence aiding detection and classification of gastrointestinal image. This competition is divided into four subtasks, including detection, efficient detection, efficient detection (the same hardware for all the participants) and report generation. We participate in the multi-label detection task and propose a new attention model, which can effectively improve the network's ability to classify different types of categories. Our approach also uses a series of different techniques including multi-epoch fusion, automatic data augmentation selection, and adaptive threshold selection. Combining these techniques, we are able to achieve good classification results on the given dataset. Finally, our f1 score is 0.907 and MCC is 0.952 with a high speed.

Probabilistic topic models are widely used to discover potential topics in a collection of documents, while latent feature vector representations have been used to achieve high performance in many NLP tasks. In this paper, we first make document topic vector representations by combining LDA and Topic2Vec, and then we perform document representations based on the topic vectors and the document vectors obtained through Doc2Vec training. Experimental results show that our new model has produced significant improvements in topic consistency and document classification tasks.

With the increasing complexity of Internet applications, traditional software architectures have been unable to support the pressure of system access brought about by user growth. Distributed systems have gradually become the mainstream architecture, and messaging has become a widely adopted model. Akka is a distributed framework based on the Actor message communication model. At present, the fault and anomaly detection for the Actor system is mainly to capture the anomaly in the code writing, it is difficult to decouple from the program, so an algorithm using kernel principal component analysis algorithm based on message monitoring is proposed to detect anomaly on Actor system. In this paper, we obtain the message of Actor system by using AspectJ's slicing of the byte code injection of Java code, and we can use Kernel Principal Component Analysis algorithm to perform data dimension reduction and feature extraction through nonlinear mapping. Then the k-means algorithm was used for cluster analysis. The LOF (local outlier points factor) algorithm was used to compare the density of each point p and its neighborhood points to determine abnormal points. Finally, we took the spider program based on the Actor model as a case to collect data and do the experiment, which verified the validity and rationality of the method.

How to improve iOS sliding performance has always been the focus of iOS application optimization. This paper analyzes the principle of AutoLayout and Frame view layout, the opportunity of network loading, CPU and GPU performance consumption during sliding process. First, we provide the appropriate solution to avoid using AutoLayout, and adjust the time of network loading by preloading to reduce the waiting time dynamically. Pre-cache and asynchronous rendering to reduce the main thread CPU consumption is implemented to reduce the main thread CPU consumption, and at the same time, GPU consumption is reduced by asynchronous rendering. Finally, verify the feasibility and effectiveness of the optimization scheme by experiments. It is verified that the percentage of the main thread CPU consumption decreases by 17.2

The tracking accuracy of the traditional CamShift moving target tracking algorithm will drastically reduce when the target is occluded or the color between the target and background is similar, moreover, there may even be serious problems with missing target. To solve the problem, this paper proposes an improved algorithm based on FAST-SIFT feature detection matching. The algorithm firstly needs to manually select the tracking target as the target template, and then compares the target search window obtained by the traditional CamShift algorithm with the target template for color histogram comparison. If the Bhattacharyya coefficient between the target candidate area and the target template is greater than the detection threshold, the target tracking fails. Then, the FAST-SIFT feature detection matching algorithm is used to match the previously obtained target template with the current frame image, which re-determines the position information of the lost target. Therefore, the target can be further tracked by the traditional CamShift algorithm. Simulation experiments show that the improved algorithm can still guarantee a good tracking effect in a relatively complicated background environment.

WeChat is one of the most popular instant messaging applications in the world. It has now become an important access to variety business systems for billions of users. The vast majority of companies want to provide their business services onto WeChat in order to gain advantage in fierce market competitions. However, as far as we know, today it is not easy to access WeChat with business service. In this paper, we propose a framework to integrate business services and WeChat. On the basis of this framework, companies or entrepreneurs can provide their business services on WeChat easily. Finally, we use a case study to demonstrate how our service can be used in helping tickets sells and statistical analysis.

提出采用可视化组件技术作为播控系统平台控制端任务配置基础工具的方案,并针对多媒体任务设计相应的可视化组件,提出和改进基于多媒体任务组件的资源调度算法.由于Web应用在跨平台、版本迭代等方面有着传统客户端所没有的优势,且组件化方案可大大降低人力成本,提高任务配置效率,因此,提出的技术方案能够在不影响原有播控系统终端播放功能的基础上,提高中央控制端的灵活性、高效性和可拓展性.

This paper implements a new text region recognition algorithm that can accurately localize image text regions in natural image with complex background. The method is mainly based on the anchor mechanism of the faster R-CNN, taking into account the special features of the text area relative to other object detect tasks, so as to convert the text area detection task in the image into a general object detection task for the small area text. In this way, we can detect the text proposal directly in the convolutional feature map of the neural network, and it can simultaneously predict the text/non- text score of the proposal and the coordinates of each proposal in the image. Then we propose a text line construction algorithm that can combine the text regions into complete text line blocks, thus greatly improving the accuracy and reliability of our text detection model. Our text detector also works accurately in multi-scale and multi-lingual text detection tasks. It achieves 0.86 F-measure and 0.78 F-measure on the ICDAR 2011 and ICDAR 2013 benchmarks, which also confirms the accuracy of our model.

互联网对传统行业的渗透,使得O2O应用层出不穷.服务行业中,作为O2O应用的核心部分,订单管理连接了用户、商家、资金流、信息流.结合O2O服务常见的用户需求,本文设计了一种较为普适的订单处理协作方案,包括订单流的状态流转,订单自动分配以及订单管理系统分层技术架构.该设计针对服务型O2O应用可拓展性较强,实际应用表明,该设计行之有效.

提出一种Web Worker池动态管理方案,本方案使开发者无需在程序运行之前确定最优的Web Worker运行数量,而是使程序在运行时自动调整Web Worker数量,从而保证程序的最优性能.该方案解决了运行多线程Web应用时需要提前获知运行环境信息的缺点,同时也解决运行时环境不确定性带来的问题.

With the rapid development of science and technology and the growing popularity of computer networks, the scale of network users is gradually expanding, and the behavior of network users is becoming more and more complicated. A large number of studies show that the user's actual interest is closely related to the browsing behavior on the web page. Through the user browsing behavior analysis can obtain the user interest information, and then build the user interest model, so that the search results closer to the user's expectations. This paper mainly introduces the method of web log mining, which can discover the mode of web pages by digging web log records. By analyzing and exploring the rules of web log records, we can identify the potential customers of the website and improve the quality of information services to users. In the stage of user behavior analysis, this paper explores the differences in user browsing behavior in different types of access events, and calculates the user's interest based on the M5 model tree to analyze the analytic events.

Software Defined Network (SDN) is a programmable network structure which achieves separation of control and forwarding via a centralized controller. Existing cloud data center network introduces SDN architecture, and enhances data center scalability via advantages of programmable thinking and centralized control. OpenStack is a free and open-source software platform for cloud computing, which provides computing, storage and network management services for cloud computing resources. However, with rapid development of various types of services, the current SDN data center based on OpenStack cannot satisfy multiple users' service requirements. Cloud data center network with single control node cannot achieve tenants' customization of their own network. Thus, multi-tenant virtual network customization (MVNC) mechanism is presented in this paper. The tenant isolation is achieved in cloud data center network through network slice. At the same time, centralized control of tenant virtual network provides a global topology management. Tenant isolation, link switching and efficiency of flow table sending are tested in this paper. Results show that proposed mechanism can flexibly and efficiently help tenants for network customization operation in OpenStack cloud platform.

This paper presents an innovative solution in which the Cocos2d-x game engine will be applied to live APPs. As a game engine, Cocos2d-x has a set of powerful animation API. Moreover, It supports scripting in JavaScript, which makes the animation code cross-platform and possible to achieve hot update.

Wi-Fi是一种短程无线传输技术,能够在一定范围内支持互联网接入的无线电信号.它的最大优点是传输速度较高,在信号较弱或有干扰的情况下,带宽可调整,有效地保障了网络的稳定性和可靠性.另外它的传输有效距离也很长,在开放性区域通信距离可达305米,使用特殊的天线技术可以达到1000米左右,在封闭性区域通信距离为76米～122米,加入功率放大电路可以增加其传输距离.近年来,Wi-Fi技术在智能家居、工业控制、移动手持设备等嵌入式环境中需求日益增多.

通过对Hadoop NameNode高可用性的研究,探讨了5种高可用性的解决方案,并分析了各个方案的优缺点及应用场景,为Hadoop的使用厂商和开发者提供参考,对于Hadoop的推广和使用有着重要的意义.

Internet testing system is a distributed network performance monitoring system, through the perception of users to obtain the user's"real experience" information, to provide the most direct and scientifi c basis for the maintenance of the network maintenance personnel, probe system which is introduced in this paper is to simulate user's requests to gather data.

随着宽带无线接入技术和移动终端技术的飞速发展,用户逐渐形成用手机访问互联网的习惯.但由于传统互联Web网站的界面和风格是为PC屏幕设计的,用户直接在移动端使用会带来极大的不便.因此,传统互联网站的所有者不得不花费更多的成本来开发专门的手机网站,并需要额外的服务器和网络条件.本文将作为针对网页信息转换为移动端网页的首要研究内容,在网页信息提取阶段,就以面向移动端适配的原则为基准,进行取舍、转换、压缩、修改等操作,以提取出适合移动端直接使用的网页信息.此研究可以帮助企业和个人开发者大大减少工作量,高效率地开发移动端页面.

提出一种微信公众账号代维平台系统的设计,以求最大程度地利用好微信营销手段.介绍了系统的实现目标,提出了一套包括系统总体结构方案、系统工作流程设计在内的设计方案,并对未来的平台发展进行展望,从而实现更好发挥微信公众账号营销作用的目的.

Heritrix是一个开源的、可扩展的Web爬虫,页面下载功能强大,但Heritrix被定义为一个通用网络爬虫,其在增量爬取方面功能有限.针对目前使用较为广泛的两个Heritrix版本,首先讨论Heritrix 1.14.4增量功能模块的不足之处,然后基于Heritrix 3.x,详细阐述一套新的增量功能框架的设计方案,这个框架既能有效利用Heritrix 3.×的页面下载功能,又能高效地对数据进行增量抓取.

随着互联网产业的大力发展,在全业务运营的格局下,互联网业务保障是保证竞争力的关键,在提升互联网网络质量的同时,更要重视提升互联网用户满意度.本文提出了一种互联网业务质量拨测系统,用以获取用户真实体验信息.介绍了该系统的实现目标,并提出了包括系统总体结构方案、系统硬件架构和系统软件架构设计在内的一整套设计方案,可以实现对网络性能进行拨测、及时告警和提供性能分析报告等功能.