| PORNSOMCHAI SORAPON | M, 2回目発表 | サイバネティクス・リアリティ工学 | 清川 清, 和田 隆広, 内山 英昭, 磯山 直也 |
|
title: Kinesthetic Haptic Feedback with Combined Soft Actuators
abstract: Kinesthetic sensation has an important role for immersive experience in virtual environment. Recreating this sensation for virtual interaction has been actively research on. Here we purpose a possible solution with combination of two types of actuators that can create wide range of Kinesthetic sensation which would be difficult to recreate with only one. Utilizing soft material we created lightweight and flexible suit which takes advantages of shape adaptability and compatibility with human body resulted in more comfortable and fatigue reduction for long period of usage. language of the presentation: English | |||
| 川瀬 寛也 | M, 2回目発表 | インタラクティブメディア設計学 | 加藤 博一, 清川 清, 神原 誠之, 藤本 雄一郎, 澤邊 太志 |
|
title: Touch-Typable Text Entry Systems for Augmented Reality in Everyday Life
abstract: Augmented reality is becoming popular through hand-held devices, but not through head-mounted displays. One of the reasons for this is that it is difficult to entry text. To solve this problem, it is necessary to implement a system that allows comfortable text entry without being limited by the situation or posture, as is the case with currently popular devices. In this study, we propose a text input system that enables touch typing using a smartwatch as an input device for augmented reality environment. Through preliminary experiments, we established the layout and operation method. language of the presentation: Japanese 論文題目: 日常利用の拡張現実感環境におけるタッチタイピング可能な文字入力システム 論文概要: 拡張現実感はハンドヘルドデバイスによって一般に普及しつつあるが、ヘッドマウントディスプレイによるものは日常利用までに至っていない。その理由の1つに文字入力のしづらさがあると考える。これを解決する為、現在普及しているデバイスと同様、状況や姿勢に制限されず快適に文字が入力できるシステムを実装する必要がある。本研究では、拡張現実感環境においてスマートウォッチを入力装置としたタッチタイピング可能な文字入力システムを提案する。予備実験を通してレイアウトや操作方法を確立した。 | |||
| MUHAMAD ALDY BINTANG | M, 1回目発表 | インタラクティブメディア設計学 | 加藤 博一, 安本 慶一, 神原 誠之, 藤本 雄一郎, 澤邊 太志 |
|
title: *** Color used Representation using Augmented Reality for Data Monitoring in IoT-based System *** abstract: *** IoT-based system has ability to bring the physical condition using sensors and transmit the digital data via internet in order to be analyzed afterward. However, one of the challenges, in context where encounter with numerous similar type of physical environment is how to distinguish one part from another. Augmented reality (AR) able to show the information seamlessly by visualize it overlay the physical object, which it able to tell IoT devices’ location differences in 3D space. Showing the IoT information, such as temperature or electrical current, in AR view usually deal with numerical data and texts, which can lead into difficulty how to present them. It’s also concerning on how the user can see them from where they stand. We proposed our system to show them by using color representation, as this method is very useful to show relationship one data to each other without prior knowledge. We want to create a realization of the real-time data from IoT-based system with using color overlay on-site of the physical system. With this, the user able to decide quick action before analyze them from actual numerical data. *** language of the presentation: *** English *** | |||
| BASHIR HAMZA | M, 1回目発表 | インタラクティブメディア設計学 | 加藤 博一, 向川 康博, 神原 誠之, 藤本 雄一郎, 澤邊 太志 |
|
title: *** Automated Evaluation of Human Actions while using AR Task Support System *** abstract: ***AR Task Support system helps user or new workers to complete some task by providing virtual instructions through images or videos. AR task support systems are very helpfull in many tasks like training the worker, assembly and maintenance tasks but,when users perform actions using AR guidance there is some room for errors and arises the questions like 1 User is completing the subtasks or leaving it incomple? 2.User has completed the subtask correctly? 3. User has completed the task incorrectly? so, the purpose of the Research is to evaluate users actions using deep learning. *** language of the presentation: *** English *** | |||
| PHAN-UDOM PURIT | M, 2回目発表 | ソフトウェア設計学 | 飯田 元, 松本 健一, 市川 昊平, 高橋 慧智 |
|
title: An Interactive Chatbot for Compiler Error Message - A User Study on Interaction and Perception
abstract: A key hurdle that student faces when first learning how to program is difficulty in understanding compiler error messages. With the rise of bots that can support developers in various software engineering scenarios, we believe that there is an opportunity for conversational bots to assist students in diagnosing and resolving their compiler error message. In this user study, we set out to discover how the students interact with a chatbot that is designed to support them comprehending the error message and debugging their code. Interestingly, we find that our participants treated the chatbot as a search tool. Moreover, our results indicate that the user interaction spans across a wide range of queries from conceptual programming knowledge to specific compiler error messages. In addition to this, we find how participants are satisifed with the chatbot in terms of functionality and user experience. language of the presentation: English | |||
| LI GUOQING | M, 2回目発表 | ソフトウェア設計学 | 飯田 元, 太田 淳, 藤川 和利, 市川 昊平, 高橋 慧智 |
|
Title: The Convergence of Container and Traditional Virtualization: Strengths and Limitations
Abstract: Virtual Machines (VMs) are used extensively in the cloud. The underlying hypervisors allow hardware resources to be split into multiple virtual units which enables server consolidation, fault containment and resource management. However, traditional VM architecture introduce heavy overhead and reduce application performance. Containers are becoming popular options, yet such a solution raises security concerns due to weaker isolation. We are at the point of container and traditional virtualization convergence where lightweight hypervisors are integrated into the container ecosystem to incoprate both VM isolation and container performance. However, there has been no comprehensive comparison among different convergence architectures. To identify limitations and best fit use cases, we investigate the characteristics of Docker, Kata, gVisor, Firecracker and QEMU/KVM by measuring the performance of disk storage, main memory, CPU, network, system call, and startup time. QEMU/KVM is used as an example of traditional VMs, Docker as the standard runc container, and the rest as the representatives of lightweight hypervisors. On top of all, we review the whole spectrum of virtualization technologies, starting from the first virtual memory implemented in IBM 370 system to a widely adopted container orchestration system Kubernetes. We study how virtualization is achieved by trap and emulate in the 1970s and how x86 ISA limitation is addressed using binary translation. To complete the picture of modern virtualization, we examine the underlying kernel features: cgroups and namespaces that enable process and resource isolation. we compare and analyze the benchmark results, discuss the possible implications, explain the trade-off each organization made, and elaborate on the pros and cons of each architecture. For this particular presentation, I will cover a case study of KubeVirt - an extension of Kubernetes that enables running VMs along side with containers, explain our findings related to CPU pinning and NUMA pass-through and how we can achieve optimal performance with vairous tuning practices. Language of the presentation: English | |||
| THONGLEK KUNDJANASITH | D, 中間発表 | ソフトウェア設計学 | 飯田 元, 藤川 和利, 市川 昊平, 高橋 慧智 |
|
Title: Federated Learning Infrastructure for Collaborative Machine Learning on Heterogeneous Environments
Abstract: Open collaboration has been vital for the rapid and successful growth of the software industry. Software development infrastructures, such as GitHub for source codes and DockHub for container images, allow individuals from diverse backgrounds and organizations to work together and create complex and large-scale software that even big tech companies find it challenging to develop and maintain. However, such collaborative infrastructure is not yet available for machine learning models. This fact has led to the current situation where state-of-the-art machine learning models can only be built by a handful of big information technology companies that have access to gigantic datasets and massive amounts of computing resources. Inspired by the success of software development infrastructures, an infrastructure for collaboratively developing machine learning models could allow researchers to work together and potentially build better models than big companies can. The simplest approach for building a machine learning model by collective effort is to aggregate the datasets contributed by participants into a single repository and train a model using the aggregated datasets. This approach is often infeasible with privacy-sensitive data, because data privacy, security, and ownership prevent institutions from sharing their data with others. As a result, machine learning models built by open collaboration were trained only using public open datasets, and their size and diversity were inherently limited. A mechanism to aggregate diverse computing resources is also needed to build a single model in a distributed manner. To provide an environment for building machine learning models in a collaborative manner, an infrastructure that allows individuals to participate in the development and improvement of a machine learning model using their computing resources and private datasets is needed. Federated learning is a promising approach that trains a single global model from datasets distributed across many clients; however, it assumes that a single common model can be shared by all participants. In this research, I will design an infrastructure for collaborative machine learning based on federated learning that allows users to share their diverse models with others without directly sharing their datasets. Language of the presentation: English | |||
| KANG XINGYUAN | M, 1回目発表 | ソフトウェア設計学 | 飯田 元, 藤川 和利, 市川 昊平 |
|
Title: Distributed SDN Controller Abstract: After SANE, Ethane, OpenFlow, the concept of SDN was introduced. SDN is a new network architecture, whose core concepts are separation of control and forwarding plane, centralized control of network state, and support for software programming. It is mainly divided into application plane, control plane, and infrastructure plane. From the definition of SDN network architecture, I find that the controller in the control plane is an extremely important existence. The controller is like a brain that controls the whole network architecture. In this colloquium, I will focus on the importance of the controller to explain the potential problems in the controller. Faced with these problems, I have four solutions and I'll explain my solutions in detail in this colloquium. Language of the presentation: English | |||
| 浅田 尚紀 | D, 中間発表 | 計算システムズ生物学 | 金谷 重彦, 作村 諭一, 小野 直亮, MD.ALTAF-UL-AMIN, 黄 銘 |
|
title: Evaluation of intercellular lipid lamellae in the stratum corneum by polarized microscopy
abstract: 角層の細胞間脂質は、ラメラ構造を構築し、外部刺激から生体を守るバリア機能を担っている。これまで報告されている細胞間脂質の評価法は、煩雑な操作が必要か、または侵襲性の高いものであった。そのため、本研究では細胞間脂質含有量を簡単かつ迅速に評価する方法の開発を試みた。ラメラ構造は偏光顕微鏡で白く観察されることから、角質層の偏光画像から計算された輝度値から細胞間脂質含有量を推定できると考えた。 偏光画像の輝度と、細胞間脂質の主成分のひとつであるセラミドの量との関係を、Image Jおよび畳み込みニューラルネットワーク（CNN）よる評価モデル構築で解析したところ、相関がみられた。 この方法では、多数の検体の細胞間脂質を効率的に評価する被験者から剥離角層を提供してもらうことで迅速に評価結果を提供することができる。 language of the presentation: Japanese | |||