| 朝比奈 甲樹 | M, 1回目発表 | コンピューティング・アーキテクチャ | 中島 康彦, 林 優一, 張 任遠, KAN Yirong, PHAM HOAI LUAN |
|
title: Accelerate Graph Attention Networks using IMAX
abstract: Graph Attention Network (GAT) is a machine learning model used in a wide range of fields, including shopping site recommendation systems, chemical reaction prediction, and natural language processing. However, like existing machine learning models, GATs require high computational power, making them power inefficient when using GPUs and other devices. To solve this problem, I will improve power efficiency by using IMAX, a Coarse-Grain Reconfigurable Architecture (CGRA). language of the presentation: Japanese | |||
| 藤原 和真 | M, 1回目発表 | 光メディアインタフェース | 向川 康博, 林 優一, 舩冨 卓哉, 藤村 友貴, 北野 和哉 |
|
title: Proposal of Coded Imaging Technology for Identifying Diverse Optical Properties
abstract: In this study, we propose a new method for identifying diverse optical properties using coded imaging technology. This technique encodes spectral information of light by minutely varying camera parameters over time, and decodes it through computational processing to identify specific patterns. This approach is expected to enable high-precision and efficient identification under low-light conditions, which was challenging with conventional methods. We verified the principle of this method through simulation experiments and demonstrated the feasibility of this method. language of the presentation: Japanese 発表題目: 多様な光学特性を識別する符号化撮像技術の提案 発表概要: 本研究では, 符号化撮像技術を用い, 多様な光学特性を識別する新しい手法を提案する. この手法は, 光のスペクトル情報をカメラパラメータを時間的に微小量変化させることにより符号化し, 計算処理により復号化することで特定のパターンを識別することができる技術である. このアプローチにより, 従来の手法では困難であった低照度での識別が, 高精度かつ効率的に行えることが期待される. シミュレーション実験により本手法の原理検証を行い, その実現可能性を示した． | |||
| 中村 海翔 | 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:AR training system for precision task using an event camera and a high-speed projector
abstract:In recent years, augmented reality (AR) technology has begun to be applied as a support and training system for tasks such as assembly and inspection work. In task training, there are two main categories of things that people need to understand and acquire: "procedures" that show the flow of the task and "tricks" to perform the task well. In this research, we focus on the acquisition of "tricks," which are relatively less studied, and focus on "precision tasks," for which these are particularly important. We take soldering as an example of a precision task, and investigate how the acquisition of tricks is affected when the difference between the optimal behavior and the current user state is immediately presented using AR technology. For this purpose, we develop an AR precision work training system using an event camera and a high-speed projector, and verify its effectiveness. language of the presentation:Japanese 発表題目:精密作業におけるコツの習得のためのイベントカメラと高速プロジェクタによるAR訓練システム 発表概要:近年，拡張現実感(AR)技術が，組み立てや点検作業などの作業の支援・訓練システムとして実応用され始めている．人が作業訓練において，理解し，身につけるべき事柄は，作業の流れを示す「手続き」と，作業をうまくこなすための「コツ」に大別されるが，現在の多くの研究は前者の記憶の効率化に焦点を当てている．本研究では，比較的検討の少ないコツの習得に着目し，それが特に重要な「精密作業」に焦点を当てる．精密作業の一例としてはんだ付け作業を取り上げ，AR技術により，最適な振る舞いと現在のユーザの状態の差を即座に情報提示した場合，コツの習得にどのような影響を与えるかを調査する．そのために，イベントカメラと高速プロジェクタを用いたAR精密作業訓練システムを開発し，その有効性を検証する． | |||
| 的場 悠希 | 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, 和田 隆広, 神原 誠之, 藤本 雄一郎, 澤邊 太志 |
|
title:Data Collection System for Autonomous Vehicle Stress Estimation Considering Object of the Passenger's Gaze
abstract:For the continued use of automated vehicles, it is necessary to estimate the factors of autonomous vehicle stress felt by passengers in real-time, and to reduce the stress. Real-time estimation of stress factors can be achieved by learning the relationship between driving environment data and passenger stress information, but the only method for collecting driving environment data including passenger gaze point information and biometric information exists using manual annotation, and it is difficult to create a huge data set for learning. In this research, I propose a framework for automatically creating a data set consisting of information on the vehicle's own behavior, the vehicle's external environment, the passenger's biometric information, and the passenger's gaze point for estimating factors that cause stress in autonomous driving, using an omnidirectional camera, which has been widely used as an in-vehicle camera in recent years. I also investigate the effectiveness of the proposed framework by creating an estimator of passenger stress factors using the dataset created by the proposed framework and verifying whether it is possible to estimate the factors of autonomous vehicle stress. language of the presentation:Japanese | |||
| 波越 貴也 | M, 1回目発表 | サイバネティクス・リアリティ工学 | 清川 清, 加藤 博一, 内山 英昭, Perusquia Hernandez Monica, 平尾 悠太朗 |
|
title: Improving interaction of Pseudo-Haptics with Attention Manipulation
abstract: Pseudo-Haptics is a representative tactile presentation technology that utilizes cross-modal perception in a virtual space. Pseudo-Haptics is a phenomenon in which pseudo-haptics is generated by applying appropriate changes in position and velocity to visual stimuli that reflect body movements. However, Pseudo-Haptics for interaction in a virtual space causes a gap between the actual motion and the motion in the virtual space that reflects it, resulting in spatial inconsistency. On the other hand, when the effect of haptic-visual interaction is generated in real-world interaction, it is necessary to ensure spatial consistency with the real world so that the interaction with the real-world environment is not disturbed. Therefore, this study aimed to improve the interaction of Pseudo-Haptics by using attentional manipulation to make the user look away from the object where spatial inconsistency occurs and correct the misalignment without the user noticing. language of the presentation: Japanese 発表題目: 注意操作を用いたPseudo-Hapticsのインタラクション向上 発表概要: 仮想空間内で、クロスモーダル知覚を活用した触覚提示技術の代表としてPseudo-Hapticsがある。Pseudo-Hapticsは身体動作を反映した視覚刺激に対して、位置や速度に適切な変化を与えることで擬似触力覚が生じる現象である。 しかし、仮想空間内でのインタラクションを対象としたPseudo-Hapticsは実際の動きとそれを反映した仮想空間上での動きの間にずれを生じさせるため、空間的不整合が生じる。一方、実世界におけるインタラクションにおいて視触覚間相互作用の効果を生起させる場合には実世界の環境とのインタラクションに支障が出ないよう，実世界との空間的整合性を担保する必要がある。 そこで本研究は、注意操作を用いて空間的不整合が発生している物体からユーザの視線を逸らさせ、ユーザに気づかせずにずれを修正することで、Pseudo-Hapticsのインタラクションを向上させることを目的とした。 | |||
| ☆HABIBA HESHAM MOHAMED ABDELAZIZ | M, 1回目発表 | サイバネティクス・リアリティ工学 | 清川 清, 加藤 博一, 内山 英昭, Perusquia Hernandez Monica, 平尾 悠太朗 |
|
title: *** Locomotion Techniques in Virtual Reality ***
abstract: *** Redirected walking is a technique that enables users to navigate a large Virtual Enviroment in a limited physical tracked area. To increase the sense of prescence in the VE, Virtual objects are mapped to a physical object in the real space to provide haptic feedback in a process called Alignment. This research aims to reduce the angle error in this process as it was not considered in previous reserach *** language of the presentation: English | |||