| 西田 篤史 | 1551073: M, 1回目発表 | 視覚情報メディア |
|
title:
[Paper Introduction] OverFeat: Integrated
Recognition, Localization and Detection using Convolutional Networks
| ||
| 山本 雄貴 | 1551115: M, 1回目発表 | インタラクティブメディア設計学 |
|
title: *** An introduction to SLAM++ : Simultaneous Localization and Mapping at the Level of Objects *** abstract: *** I will introduce the paper titled "SLAM++ : Simultaneous Localization and Mapping at the Level of Objects". Visual SLAM is a method to estimate the position and orientation information of a camera in 3D from a set of sequentially taken images. Many studies used the Visual SLAM method for mapping according to low level image features such as points and lines. In contrast, SLAM++ achieves object recognition oriented SLAM. SLAM++ has several advantages 1) reduced amount of data in map creation, 2) decreased object search time in maps, 3) improved map robustness. In the presentation, I will explain key ideas of SLAM++. *** language of the presentation: *** Japanese *** | ||
| ZENG XINGJI | 1551131: M, 1回目発表 | 光メディアインタフェース |
|
title: Estimation of Subsurface Scattering Parameters based on Directional Dipole Model
abstract: Most objects around us are more or less translucent so it is meaningful to analyze translucency of objects. However, the optical phenomenon of subsurface scattering happens inside the object, hence it makes the analysis difficult. In our research, we propose a method to estimate subsurface scattering parameters from image using directional dipole model, which is a newly proposed model for subsurface scattering. We choose it because of its capability to approximate subsurface scattering effect precisely. To test the our approach, we estimates the scatterng parameters in simulation. As a result, we confirm the validity of the proposed method. language of the presentation: Japanese | ||
| 八木 英光 | 1551110: M, 1回目発表 | 環境知能学 |
|
title: An effective method to share understandings about risks of consumer systems with end-users
abstract: For safety critical systems like avionics and railway systems, some documents to share common understandings of risks among stakeholders are often required by international standards. On the other hand, recently people worry about that even consumer systems like personal care robots have unprecedented risks and thus it is expected that documents to share risks among stakeholders, including the end-user in this case, will also be required for such consumer systems. However, since in general the number of users of a consumer system is huge and use-cases can vary, using the traditional notations for documents to share risks might fail to convince end-users, which implies to fail the purpose of the document, i.e. to share common understandings of the risks. In this research, I will aim to propose a method for sharing risks about consumer systems from the view of end-users. In the presentation, I will introduce related works and show my research plan. language of the presentation: Japanese | ||
| INSISIENGMAY ALIVANH | 1551123: M, 1回目発表 | 自然言語処理学 |
|
title: Morphological Analysis For Lao language abstract: In natural language processing tasks, word segmentation and part of speech tagging are the fundamental tasks to be conducted before proceed to the high-level tasks like parsing, semantic role labeling and so on. Lao language is similar to other Asian languages like Japanese, Chinese and Thai that are written without word boundary delimiters which makes it challenging to process a word segmentation. Not to mention about the research on part of speech tagging for Lao language, there is no research on it yet and it would be an interesting task to implement to prepare resources for a high-level tasks in Lao language processing research in the future. Thus, we propose to use Deep Neural Networks which is recently becoming the powerful tools for many natural language processing tasks and well-known method of avoiding using any handcrafted features to perform joint word segmentation and part of speech tagging in single step simultaneously. language of the presentation: English | ||
| 丸山 和輝 | 1551104: M, 1回目発表 | ネットワークシステム学 |
|
title: RFID Bionavigation System abstract: This study proposes a small lesion marking system based on an RFID technique.RFID Bionavigation System provide the position of miniature RFID-tag implanted by the vicinity of the lesion. The marking system measures the distance between the sensor and RFID-tag which is placed as a marker in diagnosis stages by marking efficient use of received signal strength from the tag. language of the presentation: Japanese | ||
| 山中 大幸 | 1551113: M, 1回目発表 | 生体医用画像 |
|
title: Registration of clinical CT and micro CT in femoral head abstract: Bone Mineral Density and Trabecular structure in femoral head or neck are important information for treatment of bone fracture. The purpose is registration of clinical CT and micro CT for prediction of those information. I will solve various problems in that registration. language of the presentation: Japanese | ||
| 浅山 和宣 | 1551003: M, 1回目発表 | 数理情報学 |
|
title: Modeling of Lane Changing Intention by Using Surrounding Vehicle
abstract: Advanced Driver Assistance Systems (ADAS) can help drivers to improve traffic flow and avoid accidents. And an autonomous car system can be main transportation system in future. These are the next generation driving system. For realization of these systems, modeling of driving behavior and extracting characteristics of the road is important. In particular, predicting lane changing intension in advance can prevent accidents in situation that other vehicle is reaching. There are several researches about prediction of lane changing, but none focus on surrounding vehicles. In our study, we tried making predicting model for lane changing from surrounding vehecles information of real driving data on highway. Language of the presentation: Japanese | ||
| 石倉 裕貴 | 1551009: M, 1回目発表 | 知能システム制御 |
|
title: Sit-to-stand Timing Prediction from Surface Electromyography Signals
abstract: Estimation and prediction of user motion intention are crucial for assistive robotics. In this study, we study an approach for predicting sit-to-stand timing from surface electromyography signals. Gaussian process regression (GPR) with the feature extraction technique so-called automatic relevance determination (ARD) is utilized in our method. Experimental results suggest the effectiveness of our approach. Language of the presentation: Japanese | ||
| 尾中 哲哉 | 1551027: M, 1回目発表 | ロボティクス |
|
title: Development of robotic system of buttoning up abstract: In cleaning factories, robotic process automation is strongly expected. Manipulation of deformable objects such as clothes is difficult even with the latest technology. Especially the buttoning up of clothes is the most challenging task to be achieved. This presentation will give an overview of the manipulation and cognitive systems of deformable objects to establish the robotic system of buttoning up. Then, I will introduce a tool called the "button aid" and discuss a method which enables robots to handle this tool just like humans. language of the presentation: Japanese | ||
| 孔 惠子 | 1551041: M, 1回目発表 | 計算システムズ生物学 |
|
Title: A research on Cine-MRI image improvement with free breathing
Abstract: Cardiac MRI can evaluate heart function and diagnose ischemic cardiac disease. Cine-MRI can observe the shape and evaluate function of heart as movies by, however, there are also disadvantages of poor image quality. Generally, we can improve image quality by classifying the cardiac phase under synchronous electrocardiogram recording and performing weighted averaging the phase image. However, such approach is not applicable to the patients with arrhythmia. In this study we aim to develop a new method for classifying cardiac phases, without synchronous electrocardiogram, and improving the image quality. In this presentation I will introduce the previous related research work, and describe the proposed method. Language of the presentation: Japanese | ||
| 土田 将司 | 1551063: M, 1回目発表 | ディペンダブルシステム学 |
|
title: Mobile agent gathering problem with Byzantine failure
abstract: A mobile agent, which is a program that can move on distributed systems, has attracted a lot of attention because it simplifies design of distributed systems. As a fundamental algorithm of agents, the gathering algorithm is considered. The gathering algorithm is to collect all agents into one node. By using the gathering algorithm, agents can exchange information. My work is to solve the gathering problem under presence of Byzantine failure. Byzantine failure models a cracked agent, that makes a malicious behavior to disturb gathering. In this presentation, I introduce the existing gathering algorithm under presence of Byzantine failure and explain the direction of my future research. language of the presentation:Japanese | ||
| 中村 優吾 | 1551068: M, 1回目発表 | ユビキタスコンピューティングシステム |
|
title:Development of IFoT(Information Flow of Things) Prototype : Toward the Realization of Smart Society Services
abstract:Toward the Internet of Things era, we propose a system called IFoT (Information Flow of Things) to process information flows gushing out from enormous IoT devices and make interactions according to the situation in a real space. IFoT employs a multi-layered architecture that scales in three stages (Point, District, City). It aims to realize a mechanism to promote simultaneous utilization for different purposes of information flow. In this presentation, challenges needed to achieve IFoT is proposed and discussed. language of the presentation:Japanese | ||
| RADEVSKI STEVCHE | 1551129: M, 1回目発表 | ソフトウェア工学 |
|
title: Towards Building API Usage Examples Metrics abstract: It is not unreasonable to say that examples source code is one of the most commonly used knowledge sources when learning the usage and best practices of a new API. That being said, in many cases the examples provided on the APIs website are lacking in quantity or quality, so developers have to resort on other information sources, namely blogs and coding forums. Unfortunately, there is no good way for API developers to measure anything concerning the examples they are creating providing. In order to resolve the problem of lacking example information and feedback, our goal is the development of metrics for empirical measurement of examples and support during the APIs example creation steps, and the work until now represents the starting point towards that aim. In the current work, we have analyzed the source code examples provided on the API’s website or GitHub directory for 7 popular API libraries written in Java and measured certain metrics such as example coverage, example code to source code ratio, class coverage percentage, and problems that occur with the compilation and execution of existing examples. Our goal is to build automatic tools for measuring various metrics for source code examples, and provide a support framework for creating better and more useful examples. language of the presentation: English | ||
| 神内 良太 | 1551053: M, 1回目発表 | 情報基盤システム学 |
|
title: A study on new security threat for Industrial IoT (Internet of Things)
abstract: Hitherto, control system is safer than ICT system because it used own standard and didn't connect external network directly. But now, industrial circles accommodate IoT to control system in order to make efficient and be highly advanced. For example , German government is implementing new Industrial policy that called "Industry 4.0". These systems are based on conventional ICT system. However control systems that use in the factory has some peculiar problems from a security perspective. So,security measures about control system become more important currently. In this presentation, I will explain security thread for Industrial IoT and introduce previous study. Finally , I will show the direction of my research plan. language of the presentation: Japanese | ||
| 湯下 弘祐 | 1551117: M, 1回目発表 | インターネット工学 |
|
title:introduction of the paper" Detecting APT Malware Infections Based on Malicious DNS and Traffic Analysis"
abstract:When detecting general malware attacks, typically Signature and Anomaly Detection Methods are used. However, Targeted Attacks are more difficult to detect with such methods, since these attacks are specialized to the target. Therefore, methods other than simply Signature and Anomaly must be investigated. To that end, I will introduce existing work to detect a specific kind of Targeted Attack known as APT infection by analyzing malicious DNS and network traffic. To detect infections in the network, IDnS (a new system introduced in this paper) requires a learning period of malicious DNS queries in the data set. language of the presentation:Japanese | ||