| 松村 遥 | 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, 小笠原 司, Christian Sandor, 武富 貴史 |
|
title: Warping Facial Contours in Face Swapping Applications for Improving Self Recognition
abstract: To facilitate the effect of mental practice on sports training, we need face swapping video that has high self-recognition. In our approach, we try to solve this problem via warping facial contours. Therefore, in my research, we verify whether self-recognition is improved in face swapping video that is modified facial contours. In this presentation, I will show the research background details, the result and the consideration of the preliminary experiment, and the future schedule of this research. language of the presentation: Japanese | |||
| 髙田 将志 | M, 2回目発表 | ユビキタスコンピューティングシステム | 安本 慶一, 小笠原 司, 荒川 豊, 藤本 まなと |
|
title: Ubi-Instructor：Development of Body Weight Training Support System by Wearable Devices
abstract: In recent, chronic lack of exercise is a problem over the world. In this study, we focus on body weight training (BWT) which only uses body weight load to solve the problem. BWT is recognized as an exercise that can be easily practiced using the load of own body weight only, and is a practical and simple training method that can train muscles of the whole body around the core of body muscle. The feedback of form, proper training menu recommendation and continuous is important for maximizing the effect of BWT. However, unlike BWT under the guidance of a personal trainer, it is difficult to achieve the above mentioned effects when exercising alone, especially for amateurs and beginners. The objective of this study is to realize a novel support system which allows beginners to perform effective BWT alone, with wearable devices. We clarify proper sensor position which achieves high recognition BWT types accuracy and propose BWT types recognition method by machine learning to realize the system. language of the presentation: *** Japanese (choose one) *** | |||
| 脇本 宏平 | M, 2回目発表 | 知能コミュニケーション | 中村 哲, 小笠原 司, 吉野 幸一郎 |
|
title: End-to-end learning of actions and instructions of life support robot
abstract: It is expected that the life support robot can explain its actions in natural language and generate actions from instruction sentences in natural language. For this mutual learning, we propose a model that learns relations between natural language and robot states such as joints trajectory and images from a camera of robot using recurrent neural networks and convolutional neural networks. To collect dataset, we make some actions using a robot on simulator. language of the presentation: Japanese | |||
| 高橋 いつみ | D, 中間発表 | 自然言語処理学 | 松本 裕治, 中村 哲, 新保 仁, 進藤 裕之 |
|
title: Sentence normalization for noisy text
abstract: There are many kinds of texts that are written in a non-standard style such as social media texts or dialects. These texts include many lexical variants such as insertions, phonetic substitutions, abbreviations that mimic spoken language. The normalization of such a variety of non-standard tokens is one promising solution for handling noisy text. We propose two approaches for the text normalization task. In the first approach, we propose a method of incorporating character- and morphological normalization to conventional Japanese morphological analysis. In the second approach, we propose a neural network based method that normalizes sentences at the character level. language of the presentation: Japanese | |||
| 中原 英里 | M, 2回目発表 | 数理情報学 | 池田 和司, 安本 慶一, 久保 孝富, Nishanth Koganti |
|
Title:Differences in Gaze Behavior between Dog Experts and non-experts During Observation of Dog Training Videos Humans can interpret social signals from not only other humans but also other species with sociality such as domestic dogs. However, little is known about how they detect and understand social signals from other species. In this study, we investigated the gaze behavior of dog experts and non-experts while they are watching dog training movies. We measured the gaze behavior of participants watching these videos using an eye-tracker. We then clipped some scenes which are acted by dog-trainer. We defined regions of interest (ROI) on a dog and a trainer. Then, we counted gaze points in ROI during observing these movies, and compared differences in gaze points between experts and non-experts. As a result, dog experts gazed the dog face more times frequently than non-experts during the time segment from commands to assessments. Language of the presentation: Japanese | |||
| 久田 将史 | M, 2回目発表 | 数理情報学 | 池田 和司, 安本 慶一, 久保 孝富, 佐々木 博昭 |
|
TItle: Analysis of Drivers' Gaze Behavior on Winding Roads.
Abstract: Drivers’ gaze modeling is important to understand steering behaviors and has been well studied in simulated/natural situations. However, many of them only analyzed the gaze in a single curve and few discussed the continuous curves. To see a conventional model called the tangent point (TP) model works even in the continuous curves, we collected real gaze data during driving on winding roads and analyzed their properties. As a result, the vertical position of the gaze did not correlate to the time while the horizontal position did, and drivers’ eye movements were not saccades but smooth pursuits, indicating that the gaze did not jump from one TP to the next one. Language of the presentation: Japanese | |||
| M. ROSYIDI | M, 2回目発表 | 数理情報学 | 池田 和司, 安本 慶一, 久保 孝富, 佐々木 博昭, Nishanth Koganti |
|
title: *** A Design of IoT-based Searching System for Displaying Victim’s Presence Area
***
abstract: *** As a disaster is an unpredictable event, it may be tough to escape damage entirely even if evacuation drills are carried out periodically. In such a case, persons requiring evacuation assistance like aged people and disabled people are prone to be victims. For minimizing the damage, a fire departement and a local government need to swiftly grasp the situation of the afflicted area in order to execute rescue operation. In this research, to support the rescue operation, we propose a design and implementation of an Internet of Things (IoT) device to grasp the presence of persons requiring evacuation assistance. The IoT device captures Wi-Fi signal and packet from smartphones that victims have, and it then provides an estimated area for a presence of them. In the estimation of the area, we introduce three ways to select center of an area, i.e., random, average, and median selections. Through the experimental results, we showed that median selections make more accurate presence area than average selection and random selection. Also, the system presented the visualization of the area to support the rescue operation. Keywords : IoT, Wi-Fi, GPS, Presence area, Smartphones, Disaster *** language of the presentation: *** English *** | |||
| NDIBWILE JEMA DAVID | D, 中間発表 | サイバーレジリエンス構成学 | 門林 雄基, 林 優一, 藤川 和利 |