-
Ryuichi SHIBATA, Wataru HASHIMOTO, Yasuharu MIZUTANI, Satoshi NISHIGUC ...
Session ID: 21-01-01
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this research, we present a three-dimensional image of a Pepper's Ghost, and simultaneously observed from
multiple angles by rotating it. Pepper's Ghost is a method to make an image appear to be floating in the air by combining an image reflected on a transparent board with a background. The system consists of a monitor lying on its side, a transparent board with a viewing angle control film on it, and a motor. Evaluation experiments have shown that there is a trade-off between the clarity of the 3D image and eye strain depending on the rotation speed. The purpose of this study is to investigate whether the size of the transparent plate improves or increases visibility. Visibility in this study is defined as the resolution and frame rate of the stereoscopic image and the vertical viewing angle.
View full abstract
-
Tomoya YAMASHITA, Wataru HASHIMOTO, Satoshi NISHIGUCH, Yasuharu MIZUTA ...
Session ID: 21-01-02
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this study, we attempt to be able to sense the visual self-motion (Vection) to the user who utilizes our car driving system. Vection is an illusion which is caused by visual information, so a human has an illusion that he/she is moving. In this research, we make the user sense Vection by processing such as transforming and moving the rendering area when the car turns. Vection is dependent on the range of vision, so we aim to be felt Vection more effective by using Immersive Video System such as a spherical display and a head-mounted display, because of covering the user's gaze. In the experimental setup, we consider that we present senses of acceleration and deceleration by enlargement and reduction, not only horizontal movement.
View full abstract
-
-For DIY digitization of small format films-
Kazuyuki BABA
Session ID: 21-01-03
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In addition to 35mm film for theatrical use, there are many other formats of motion picture film. For example, 8mm and other small formats were popular for home use. These films are a valuable source of information about the old days. However, since these films are not managed by experts, these are often deteriorating with age. In addition, many damaged films may occur at once due to submersion in water caused by a disaster.
Digitization is considered an effective way to deal with such crises. However, film scanners that can scan deteriorating films are expensive devices. To solve this problem, I have devised an inexpensive film scanning system that can scan deteriorated films using a smartphone camera application and the web application I developed.
View full abstract
-
Junpei HAYASHI, Yota YAMAMOTO, Yukinobu TANIGUCHI
Session ID: 21-01-04
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In industry such as manufacturing or restaurant, many companies use videos of experts’ work to efficiently transfer their skills to unskilled workers. However, long videos decrease the efficiency of skill transmission because it takes long time to watch the videos. Although scene retrieval in the videos increases efficiency, it requires dividing the videos by work process. Instructional videos usually contain less visual changes compared to usual, and previous scene segmentation methods that depend only on appearance changes often fail to divide instruction videos appropriately. In this study, we propose a scene segmentation method that utilizes sentence information about work processes such as instruction manuals.
View full abstract
-
Koki NISHIMOTO, Kimiaki SHIRAHAMA
Session ID: 21-01-05
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
When a video is played back, humans understand the contents while unconsciously deciding whether to memorize each frame. This paper presents a method to acquire this human memory mechanism using reinforcement learning. By defining a memory as a finite external storage, our method trains an agent that can update the memory so as to achieve its optimal matching with frames observed over time. Inspired by imitation learning that learns a reward function from expert’s demonstrations, the agent gets a reward representing how well a frame matches the memory. The effectiveness of the proposed method is demonstrated through the experiments on video data provided by TRECVID.
View full abstract
-
Yuta Okuda, Yota Yamamoto, Yukinobu Taniguchi
Session ID: 21-01-06
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
With the aging of dairy farmers and lack of successors, dairy farming is becoming more intensive on a large scale. In order to detect signs of estrus and disease in early stages, dairy farmers need to observe each cow carefully, but the burden is increasing due to a lack of manpower. In this paper, we propose an image processing method for detecting mounting behavior, which is important for detecting signs of estrus. Specifically, we propose a method for detecting abnormal animal behavior (mounting) by inputting the image of a detected dairy cow cropped in the center into an abnormality detection model called GANomaly.
View full abstract
-
-What is the Free Energy Principle?
Toshio INUI
Session ID: 21-01-07
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
The various functions of animals, including humans, are considered to be designed to minimize Helmholtz free energy (or more precisely, its information-theoretic version, variational free energy). Minimizing free energy closes the cycle of perception and motion. Notably, it is through this repetition that we adapt to the world. This function is called active inference, which emerges whenever we sample information from the world by moving or acting in a way that minimizes free energy. Active inference is the proof of one’s existence. In this talk, I will introduce the process theory of the free energy principle, especially with regard to perception, motor control, and attentional functions.
View full abstract
-
Xu WANG, Makoto FUJISAWA, Masahiko MIKAWA
Session ID: 21-01-08
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
This paper proposes a sampling algorithm for multi-size particles within an arbitrary 2D shape using a Voronoi tessellation. In order to sample as numerous particles as possible within a limited space, we offer a novel minimum porositydriven algorithm capable of optimizing the existing Voronoi tessellation method. This technique ensures no overlapping between particles while increasing the particle sampling density for physics simulation. Moreover, our method allows users to set the target particle-size distribution by a pre-designed cumulative distribution function and limits the errors to a certain degree after the optimization process. We also demonstrate the applicability of this method in physical simulations, such as the Discrete Element Method.
View full abstract
-
Takashi Uematsu, Youngha Chang, Nobuhiko Mukai
Session ID: 21-01-09
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
In this paper, we propose a method for a person to recognize corners in a virtual city only from the visual information in a walking simulation. By extracting the candidates of a corner based on the edges of a road boundary and estimating the direction of the corner from the vector direction that constructs the corner, it has become able to be possible for the person to turn the corner correctly.
View full abstract
-
Miki TACHIABANA, Motonobu KAWASHIMA, Daichi HAYAKAWA
Session ID: 21-01-10
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
There are two types of art settings for game content drawn by artists: those in which outlines and fills are completely separated, and those in which outlines and fills are mixed. However, the mixture of contour and fill, which is one of the illustration techniques, is not taken into account when creating 3DCG models from art settings in the development stage. In both cases, the visuals of the game are replaced with a toon-like or realistic look, which does not necessarily reflect the intent of the original art setting. Therefore, in this study, we proposed a rendering method to express the mixture of contour lines and paint, which is one of the important elements of art setting. As a result of practical verification, it was confirmed that the proposed method can be used to represent mixed contours and fills in 3D games.
View full abstract
-
Shuhei Munakata, Yuta Muraki, Kenichi Kobori
Session ID: 21-01-11
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
The population is increasing to learn yoga as an easy exercise. In particular, online yoga using the video call app is attracting attention as a yoga class that you can enjoy even at home. However, attending a yoga class has the problem that time is limited. Another problem is that it is difficult to receive direct guidance from the instructor due to the large number of people. In this study, we propose a system that evaluates yoga poses using skeletal estimation. In the proposed system, OpenPose, which is a skeleton estimation technology, is made three-dimensional by inputting a video from three viewpoints. In addition, a more accurate three-dimensional skeleton is generated by performing coordinate interpolation[1] using the preceding and following frames at each joint point.
View full abstract
-
Makoto J. HIRAYAMA
Session ID: 21-01-12
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Figure models of speech organs are being developed for research and education of speech production. printer is used for the development. Modeling methods are explained including previously made ones.
View full abstract
-
Steve Jobs’ Role compared with Japanese Technology and Engineers Kunio Ohno†
Kunio Ohno
Session ID: 21-01-13
Published: 2021
Released on J-STAGE: March 31, 2023
CONFERENCE PROCEEDINGS
RESTRICTED ACCESS
Human resource development and the suitable organization for the future society in view of the long−term perspective will be necessary. In this report, Steve Jobs and Alan Kay are considered as examples of key human resources, and Xerox’s PARC, which played an important role for both, were analyzed. Then, the research and development culture of Xerox, which penetrates PARC and the original electronic copy technology, has been considered which consists with author’s experience at NTT Laboratories.
View full abstract
