日時: 9月24日(月)4限 (15:10-16:40)

会場: L1

司会: 姚 駿
榮門恒希 1151022: M, 2回目発表 萩田紀博, 横矢直和, 浮田宗伯
title: System with a depth sensor for teaching skillful human motions
abstract: We propose a system for teaching skillful human motions. Common motion measurement systems are too expensive and require users to wear binding devices. On the other hand, our proposed system is possessed only of an inexpensive depth sensor. The system functionally consists of two modules below. The first one is estimating the sequence of body poses from a depth image sequence. The second one is evaluating and teaching the gaps between the estimated pose sequence and a correct one given by capturing the motions of experts. In this talk, several experimental results for pose sequence estimation would be introduced.
language of the presentation:Japanese
発表題目: 距離センサを用いた特定動作に対するティーチングシステム
発表概要: 本発表では特定動作におけるティーチングシステムについて提案する.一般的な動作計測システムは非常に高価であり,ユーザを拘束してしまうという問題がある.一方で,我々の提案手法は安価な距離センサのみを用いる.本システムは2つの機能から構成される.第一に,距離画像シーケンスから全身姿勢シーケンスを推定する機能がある.第二に,推定された姿勢シーケンスと専門家が行なった正確な姿勢データシーケンスとの差を計測,教示する機能から構成される.今回,姿勢推定機能について評価実験を行った.
上田優子 1151019: M, 2回目発表 横矢直和, 小笠原司, 佐藤智和
title:Remote Control of a mobile robot using novel viewpoint image generation
abstract:We propose the user interface of remote control of a mobile robot that users can change the viewpoint position. Proposed UI allows to grasp of the situation including the traveling direction of the robot around thereby. To realize change of viewpoint , the novel viewpoint image using environment information obtained from the omni-directional camera and the depth camera can be generated.
language of the presentation: Japanese
発表概要:本発表では操縦者による状況に応じた視点位置の変更を可能にすることで, ロボットの進行方向を含む周辺状況の把握を容易にする移動ロボットの 遠隔操縦インタフェースを提案する. また,観察視点位置の自由な変更を可能とするために, 全方位カメラと距離画像センサにより得られる環境情報を用いて自由視点画像生成を行う.
松葉有香 1151100: M, 2回目発表 池田和司, 横矢直和, 船谷浩之

title: Estimating the Clothes Characteristics with the Image and Depth Sensors for Developing Virtual Fitting Room

abstract: In this presentation, a method for obtaining the clothes characteristics by comparing the real and simulated motions gotten with image and depth sensors, Kinect and physical simulator, Blender, respectively. We use optical flow for RGB image to get the hanger trajectory and use corner detection for depth image and Dijkstra algorithm for 3D mesh to get the trajectory of real clothes. With these trajectories and simulated clothes, we optimized the parameters of clothes characteristics using a simulated annealing. 

language of the presentation: Japanese

上月一史 1151123: M, 2回目発表 加藤博一, 横矢直和, 武富貴史, 宮崎純
title: Robust feature point tracking against effects of changes in projection by considering information of camera position and orientation
abstract: AR system for estimating camera position and orientation by feature point extracted from planar texture and overlaying CG is composed. Features in this AR system are invariant with respect to pattern changes in 2 dimentional,but it's not in 3 dimentional one such as projection. Therefore, it is difficult to correspond with feature points in case of the shooting angle for planar texture is very different from registrating database. In this research, we propose robust feature point tracking by using camera position and orientation in previous frame and removing effects of changes in projection from input image.
language of the presentation: Japanese
発表題目: カメラ位置・姿勢情報を考慮した射影変化にロバストな特徴点追跡
発表概要: 平面状の紙面上のテクスチャから検出される特 徴点に基づいてカメラ位置・姿勢を推定しCG などを重 畳表示する拡張現実感技術(AR) が提案されている. このようなARに用いられている特徴量は二次元的なパターン変化に対し ては不変であるが,射影変化のような三次元的なパター ン変化には不変ではない.そのため,紙面に対する撮影 角度がデータベース登録時と大きく異なる場合は特徴点 の対応付けが難しいという問題がある. そこで,本研究では前フレームのカメラ 位置・姿勢が既知であることに着目し,入力画像に生じ ている射影変化の影響を除去することで,従来よりもロ バストな特徴点追跡を実現する.

会場: L2

司会: Kevin Duh
欅惇志 1161007: D, 中間発表 加藤博一, 松本裕治, 宮崎純, 浦西友樹
title: An Accurate XML Element Retrieval System Considering Updates of Documents
abstract: We develop an accurate XML element retrieval system handling updates of the documents. The idea of handling updates of documents is important where the document updates take place frequently, e.g., on the Web. This is because the effectiveness of a search system gets lower without it. Moreover, new topics may emerge exponentially and the statistics of the documents can change drastically, which is also harmful to search accuracy. We therefore consider a method not only for updating indices efficiently but also for processing a query effectively and efficiently. We construct indices enabling their fast updates and propose a method for computing accurate term weights even under the dynamically changing statistics. The experimental results showed that our proposed system can not only deal with document updates at low cost, but also search accurately even when the statistics of documents change.
language of the presentation: Japanese
発表題目: 文書の更新を考慮した高精度 XML 部分文書検索システム
発表概要: 文書の更新を考慮した高精度 XML 部分文書検索システムの開発を目指す.XML 文書のうちでも特に,頻繁に更新が発生する Web 文書などにおいて,データの更新を考慮しなかった場合には,やがて検索精度の低下が起こる.特定のトピックに関する文書が急激に出現し,文書集合の持つ統計量が大きく変動する場合にも,然るべき対処を行わなければ同様に検索精度が低下する.故に,Web 検索システムにおけるデータ更新への対応は重要である.そこで我々は,索引を高速に差分更新しつつ,文書の統計量が急激に変化する状況下でも索引語の重みを適切に算出することが可能となる手法の提案を行う.評価実験の結果,提案システムは低コストによる索引の差分更新を行うことが可能であり,また,統計量が変動した場合にも高精度かつ高速な検索を行うことが可能であることが明らかになった.
IGOR DE SOUZA ALMEIDA 1161014: D, 中間発表 加藤博一, 中村哲, 山本豪志朗, 宮崎純

Title: Meetsu: the use of AR Conversational Annotation in co-located electronic meeting support

Abstract: Augmented Reality (AR) has been seen as a CMC technology for its capability of augmenting the real world by annotating extra virtual information onto it. Augmenting face-to-face communication however has not been explored extensively. We introduce the concept of AR Conversational Annotation (ARCA) for co-located meeting support. This concept is based on the now popular trend of annotating over video, photo and audio as well as live annotating using tweets. This work identifies the different types of annotations alongside their current use. We propose the distinction between informative annotation and conversational annotation while discussing their use in Augmented Reality applications. Furthermore, we share our findings and insights regarding the use of AR conversational annotation for co-located meeting support based on a long-term observational study of our prototype system, Meetsu, in our weekly research meetings.

Language of the presentation: English

久保慶伍 1161006: D, 中間発表 中村哲, 松本裕治, 戸田智基, グラムニュービッグ, サクリアニサクティ
title: Investigation on automatic pronunciation annotation by using contrastive estimation
abstract: In this presentation, I discuss the pronunciation annotation of semi-supervised learning which combines CRF, which is supervised learning, with contrastive estimation which is unsupervised learning. I evaluated this method on g2p (grapheme-to-phoneme conversion) task. Through the result of the evaluation experiment, I describe the effectiveness of this method and agendas to extend this method to dictionary-based pronunciation annotation with a high accuracy for pronunciation annotation and OOV word detection.
language of the presentation: Japanese
発表題目: Contrastive estimation を用いた発音付与の検討
発表概要: 本発表では教師なし学習である contrastive estimation と教師あり学習である CRF を併用した 半教師あり学習の発音付与手法について検討する. G2p (grapheme-to-phoneme conversion) タスクにおいて本手法の評価実験を行った. 評価実験の結果を通して,本手法の有効性と,本手法を発音付与と未知語検出に関して高い精度を持つ 辞書ベースの発音付与手法へと拡張するための課題について述べる.

会場: L3

司会: 吉田 則裕
砂田隆浩 1151059: M, 2回目発表 松本健一, 飯田元, 門田暁人
title:Library Uasge popularity in Software Development of Java
abstract:In this paper, we will observe the popularity of software libraries of Java. When designing a piece of software, we must choose between external libraries. Especially in Java, we can use many middleware, framework, library and development support tools as open source software. Whereat, they must have one question. which library is the best to use ? We mined about one hundred of open source projects and analyzed the popularity of these libraries based on the idea: wisdom of the crowds.
language of the presentation:Japanese
谷宗一郎 1151067: M, 2回目発表 松本健一, 飯田元, 門田暁人, 伊原彰紀
title: Information Integration between Development Support Systems by Information Retrieval Method
abstract: The goal of my study is to make software maintenance more efficient, I propose a method to associate the similar information in each development support system automatically. Many software development projects use some individual development support system such as bug tracking system, version control system, and mailing list system. When developers note the related information in each development support system, they attach the associated tag with the information by themselves. For example, when developers change some source codes to fix bugs, they note the source codes' name in the bug report and note the bug ID in the source codes (commit message). However, many developers have not attached the associated tag. Therefore, it is difficult to track the source codes which were fixed in past times, when developer detect the similar bug. In this study, I built a system to extract the associated information from each development support system automatically. I will talk about the progress and the future work of my study.
language of the presentation: Japanese
発表題目: 情報検索手法を用いたソフトウェア開発支援システム間の情報統合
発表概要: 本研究では,ソフトウェア開発における保守作業を効率化するために,複数の開発支援システムが各々で管理する情報を自動的に関連づけることを目的とする.ソフトウェア開発プロジェクトでは,不具合票やソースコードを一元管理するバグ管理システムや版管理システムを異なる開発支援システムとして使用していることが多い.開発者は複数のシステムに関連情報を登録する場合,登録内容に手動で関連タグを付する例えば,不具合修正のためにソースコードを変更した場合,不具合票にソースコード名,及び,ソースコード(コミット記録)には不具合番号を付する.しかしながら,開発者の多くは関連タグを付しておらず,修正内容の追跡が困難となっている.本研究では,不具合票やコミット記録などのテキストに含まれる単語を用いて,各システムから自動的に関連情報を抽出するシステムを構築した.本発表では,その進捗と今後の課題について述べる.
中野大輔 1151073: M, 2回目発表 松本健一, 飯田元, 門田暁人
title: Test effort allocation strategy using a prediction module bug
abstract: In software development, are important efficiency gains in test to find bugs efficiently. To perform the test efficiency studies to predict the bug contained in (module) has been carried out many functions of the software. However, in these studies, are discussed not only to improve the accuracy of prediction, based on the prediction results or how to assign costs (labor test) Each module is not clear. Therefore, the purpose of this study was to investigate how to assign test effort to increase the rate of detection of a bug. Model the rate of bugs and test effort allocation, in this study, we performed a simulation of how much can I find a bug in the test effort allocation based on the predicted results. Summary and the way the results of simulation experiments, in this paper, we describe the future plans.
language of the presentation: Japanese
発表題目: バグモジュール予測を用いたテスト工数割り当て戦略
発表概要: ソフトウェア開発において,効率よくバグを発見するためにテストの効率化が重要視されている. テストの効率化を行うために,ソフトウェアの機能(モジュール)中に含まれるバグを予測する研究が数多く行われている. しかしながら,これらの研究では,予測精度の向上までしか議論されておらず,予測結果に基づきどのように各モジュールにコスト(テスト工数)を割り当てるかが定かではない. そこで,本研究の目的はバグ発見率を高めるテスト工数割り当て方法について明らかにする事である. 本研究では,割り当てテスト工数とバグの発見率をモデル化し,予測結果に基づいたテスト工数割り当てでどれだけのバグを発見できるかシミュレーションを行う. 本発表では,シミュレーション実験の概要と途中成果,今後の予定について述べる.
Passakorn Phannachitta 1151120: M, 2回目発表 松本健一, 飯田元, 門田暁人, 伊原彰紀
title: An Algorithm for Gradual Patch Acceptance Detection in Open Source Software Repository Mining
abstract: Nowadays, software development societies have given more precedence to Open Source Software (OSS).There is much research aimed at understanding the OSS {society} to sustain the OSS product. To lead an OSS project to a successful conclusion, researchers study how developers change source codes called patches in project repositories. In existing studies, we found an argument in the conventional patch acceptance detection procedure. It was so simplified that it omitted important cases from the analysis, and would lead researchers to wrong conclusions. In this research, we propose an algorithm to overcome the problem. To prove our algorithm, we constructed a framework and conducted two case studies. As a preliminary result, we came to a new and interesting understanding of patch activities.
language of the presentation: English