Hiroyasu Usami

Hiroyasu Usami

Chubu University

H-index: 6

Asia-Japan

About Hiroyasu Usami

Hiroyasu Usami, With an exceptional h-index of 6 and a recent h-index of 5 (since 2020), a distinguished researcher at Chubu University, specializes in the field of computer vision, medical image processing, machine learning, graph theory.

His recent articles reflect a diverse array of research interests and contributions to the field:

第 40 回ファジィシステムシンポジウム (FSS2024) 開催案内

Extraction of sample area in biomass ash sample dissolution data using time-series information

深層学習における簡素情報量を切り出す互助組織化メカニズムの確立

Blood Vessel Structure Analysis using a Simulation Model for the Purpose of Polyp Shape Recovery from Endoscopic Images

内視鏡画像からのポリープ形状復元を目的としたシミュレーションモデルによる血管構造解析

O-18 バイオマス灰の溶融試験法とその解釈に関する一考察

Seq2Seq を用いた木材に含まれる無機化合物の溶融過程の予測

Improvement of Polyp Detection using MUNIT for Image Generation

Hiroyasu Usami Information

University

Chubu University

Position

A Research Associate of Department of Computer Science Japan

Citations(all)

118

Citations(since 2020)

102

Cited By

37

hIndex(all)

6

hIndex(since 2020)

5

i10Index(all)

1

i10Index(since 2020)

1

Email

University Profile Page

Chubu University

Hiroyasu Usami Skills & Research Interests

computer vision

medical image processing

machine learning

graph theory

Top articles of Hiroyasu Usami

第 40 回ファジィシステムシンポジウム (FSS2024) 開催案内

Authors

運営組織, 実行委員長, 中村剛士, 財務委員長, 高瀬治彦, 会場担当, 早瀬光浩, 森田賢太, 荒川俊也, 岩堀祐之, 宇佐美裕康, 加藤央昌, 金久保正明, 高野敏明, 竹本修, 三好哲也, 矢野良和, 吉川大弘, 吉川雅弥, 現地外, 川中普晴, 宮本友樹, 吉田真一, 星野孝総, 堀口由貴男, 平原誠, 戸田雄一郎, 中村健二, 楠木祥文, 河辺義信, 本多克宏, 高田諒, 伊藤一也

Journal

知能と情報 (日本知能情報ファジィ学会誌)

Published Date

2024

今年のファジィシステムシンポジウムを下記の内容で開催いたします. ファジィ理論をはじめ, 知能情報システムに関する講演やイベントを予定しております. 会員の皆様からの企画提案も歓迎いたします. 多数の皆様のご参加をお待ちしております.

Extraction of sample area in biomass ash sample dissolution data using time-series information

Authors

Haruki Yamane,Hiroyasu Usami,Yuri Suzuki,Yoshihiko Ninomiya

Journal

IEICE Technical Report; IEICE Tech. Rep.

Published Date

2023/1/19

(in English) Recently, the importance of carbon neutrality has been increasing, and many biomass-fired power plants have been introduced for local production and local consumption, which are one of the decentralized power sources that form local distributed smart communities that are not affected by weather conditions.However, because the potassium content of Japanese cedar, the main thinned wood in Japan, is different from that of other thinned woods such as cypress and pine, a large amount of clinker is generated in the gasifier after a short period of operation of one to two weeks, causing plant shutdowns and other problems in a short period of time.

深層学習における簡素情報量を切り出す互助組織化メカニズムの確立

Authors

宇佐美裕康

Journal

総合工学

Published Date

2023

In this paper, as an object for studying a reciprocal architecture for simplifying deep neural networks, a method is proposed to mutually reduce parameters by combining different deep learning models and aim to construct an effective combustion prediction model of wood in gas furnaces even with a small dataset.

Blood Vessel Structure Analysis using a Simulation Model for the Purpose of Polyp Shape Recovery from Endoscopic Images

Authors

Shusuke Kato,Hiroyasu Usami,Akihiko Okazaki,Yuji Iwahori,Ogasawara Naotaka,Kunio Kasugai

Journal

IEICE Technical Report; IEICE Tech. Rep.

Published Date

2022/1

(in English) In recent years, the incidence of colorectal cancer in Japan has been on the rise. It is essential to realize a medical diagnosis support system that utilizes AI and image processing technology. In particular, in endoscopic diagnosis, the size and shape of polyps are essential guidelines for diagnosis. In order to estimate the absolute size of polyps, the Shape from Shading (SFS) method is used to restore the shape of polyps since the target object is a non-rigid body. In order to estimate the absolute size of a polyp, a shape restoration method using optical constraint equations based on Shape from Shading (SFS) has been developed since the target object is a non-rigid body. In this study, we used a reference object as a reference object to obtain the depth information from the camera to the object and the reflection coefficient C of the object. In this study, we propose a geometrical model that realizes the …

内視鏡画像からのポリープ形状復元を目的としたシミュレーションモデルによる血管構造解析

Authors

加藤俊輔, 宇佐美裕康, 岡崎明彦, 岩堀祐之, 小笠原尚高, 春日井邦夫

Journal

電子情報通信学会技術研究報告; 信学技報

Published Date

2022/1/20

抄録 (和) 近年, 日本における大腸がんの罹患率は増加傾向にあり, AI や画像処理技術を活用した医療診断支援システムの開発が望まれている. 特に内視鏡診断では, ポリープの大きさや形状が診断の重要な指針となっている. ポリープの形状を復元する際に, その絶対的な大きさを推定するために, 対象物体が非剛体であることから Shape from Shading (以下, SFS) による光学的制約式を用いた形状復元手法等が開発されている. 光学的制約式を用いる際に必要となるカメラから対象物体までの奥行情報 Z, 対象物体のもつ反射係数 C などパラメータ情報取得が必要である. 本研究では, 参照物体を血管とし, カメラと血管の構成する幾何学的モデルを用いることでそれらパラメータ算出を実現する幾何学的モデルを提案する. そこでは, 幾何学的モデルの評価は 3 次元の血管シミュレーションモデルを作成し, 形状復元に必要なカメラ座標系におけるカメラから血管までの奥行情報 Z を算出することで幾何学的モデルの評価を行いモデルの妥当性を確認している.また, 提案した幾何学的モデ …

O-18 バイオマス灰の溶融試験法とその解釈に関する一考察

Authors

二宮善彦, 佐藤龍磨, 宇佐美裕康, 星野央貴

Published Date

2022/1/12

In the combustion boilers and gasifiers of woody biomass, clinker formation and ash deposition occur in the furnace and heat recovery section, which hinders stable operation. In the case of ash with high potassium content, the generation of melt due to CaCO 3-K 2 CO 3 eutectic was confirmed in the temperature range of 750-850 C in the combustion atmosphere with CO 2 in addition to the gasification atmosphere. Since the viscosity of the melt is lower than that of the glassy melt, the melt is considered to play the role of a binder for clinker formation.

Seq2Seq を用いた木材に含まれる無機化合物の溶融過程の予測

Authors

朝日一憲, 宇佐美裕康, 岡崎明彦, 二宮善彦

Journal

電子情報通信学会技術研究報告; 信学技報

Published Date

2022/1/20

抄録 (和) 昨今, カーボンニュートラルの重要性が高まっており, 天候に左右されない地域分散型スマートコミュニ ティを形成する分散型電源の一つである地産地消型の熱電併給バイオマスガス化発電設備が注目されている. しかし, 主要間伐材であるスギ材は, ヒノキ, マツなどの他の間伐材に比べてカリウム含有率が異なるなど, その組成は異なる. そのため, 1~ 2 週間の短期間運転で炉内に大量のクリンカが発生し, 扱う木材によりガス火炉の稼働時間は異なる. 本研究では, 多様な木材の燃焼動画データに対して時系列情報を扱うモデルである PhyDNet [1] を用いることで, 扱う 木材ごとにガス火炉の最適な稼働条件を明らかにするために木材の燃焼過程の予測を行った. 本稿では 2 パターンの データセットを作成し SSIM を精度評価とし, 精度比較を行った. 結果として約 86%, 91% という結果を得ることが できたので報告する.(英) In recent years, the importance of carbon neutrality has been increasing, and biomass gasification power generation facilities with combined heat …

Improvement of Polyp Detection using MUNIT for Image Generation

Authors

Yuji Iwahori,Tsubasa Ooto,Hiroyasu Usami,Shinji Fukui,Manas Kamal Bhuyan,Aili Wang,Naotaka Ogasawara,Kunio Kasugai

Journal

Procedia Computer Science

Published Date

2022/1/1

This paper treats an image translation using Multimodal Unsupervised Image-to-Image Translation (MUNIT) from the white light source image to the Narrow Band Imaging (NBI) of the endoscope images and proposes a method to improve the detection performance by the Single Shot Multibox Detector (SSD) which trains dataset of white light source image by adding the generated NBI-like images. The proposed approach makes it possible to generate an NBI-like image by keeping the existing polyp, inner wall and specific features of coloring and brightness of the original endoscope image. It is shown that increasing the number of data of the generated images achieves the better performance. The performance of the proposed approach was evaluated using the actual endoscope images which include the polyps of the various shapes. Recall of 80.62% and precision of 93.47% were obtained as a result through the …

Behavioral analysis of preschool children to support evaluation of digital contents in AR space

Authors

Ren Ando,Hiroyasu Usami,Yuri Suzuki,Wasuke Hiiragi

Journal

IEICE Technical Report; IEICE Tech. Rep.

Published Date

2022/1/20

(in English) With the remarkable progress in the field of ICT, the introduction of ICT education using technologies such as AI, IoT, and AR has been rapidly promoted. The Komaki Children s Miraikan, a facility in Komaki City, Aichi Prefecture, provides hands-on learning using digital content. At this facility, each content are updated and improved every three months, and content developers visually check the responses status of preschool children to each content. Hence, content developers are difficult to grasp all the users reactions to the contents, and the reality is that the developers cannot fully confirm the effectiveness of the contents. Therefore, in this paper, we provide a method for detecting preschoolers on a camera and visualizing their behavior using a heat map to support effective digital content creation.

2-11 ディープラーニングを活用したフライアッシュ粒子の形状分類に関する研究

Authors

二宮善彦, 中村太一, 宇佐美裕康, 森岳人, 長沼宏, 澤田晃宏

Published Date

2022/10/13

In order to predict the trajectory of ash particles as they impact the heat transfer tubes due to inertial impact, information on particle size, density, and particle shape of ash particles is necessary. In this study, we investigated a method for shape classification of ash particles using deep learning from SEM images of fly ash particles. As a result, we succeeded in classifying ash particles ranging in size from 1 to 200 μm into eight shapes and obtaining an integrated particle size distribution for each shape.

Automatic detection of lst-type polyp by cnn using depth map

Authors

Yuji Iwahori,Shota Miyazaki,Hiroyasu Usami,MK Bhuyan,Boonserm Kijsirikul,Aili Wang,Naotaka Ogasawara,Kunio Kasugai

Journal

Handbook of Artificial Intelligence in Healthcare: Vol. 1-Advances and Applications

Published Date

2022

Lateral Spreading Tumor (LST) type flat polyps are sometimes overlooked and difficult to be detected among many lesions. This paper proposes a CNN model of multiple input and multiple output structure to detect LST-type polyp with high accuracy, which is based on U-Net architecture for the segmentation. Not only the original endoscope image but also depth map is also used to the original CNN structure of 2 inputs and 4 outputs. Here, proposed method obtains 3D shape from the original endoscope image and creates the depth map under the condition of point light source illumination and perspective projection. Higher accuracy of 85% was obtained for the detection of LST-type polyp by the proposed method. It is shown that the multiple input-output structure of U-Net model gives the higher performance of segmentation problem using both of original endoscope image and depth map.

Prediction of melting process of inorganic compounds in wood using Seq2Seq

Authors

Kazunori Asahi,Hiroyasu Usami,Akihiko Okazaki,Yoshihiko Ninomiya

Journal

IEICE Technical Report; IEICE Tech. Rep.

Published Date

2022/1/20

(in English) In recent years, the importance of carbon neutrality has been increasing, and biomass gasification power generation facilities with combined heat and power for local production and local consumption, which are one of the distributed power sources to form a regionally distributed smart community that is not affected by weather, have been attracting attention. However, the composition of cedar wood, which is the main thinned wood, differs from other thinned woods such as cypress and pine in terms of potassium content. Therefore, a large amount of clinker is generated in the furnace in a short period of time (1~ 2 weeks), and the operating time of the gas-fired furnace varies depending on the lumber handled. In this study, we used PhyDNet [1], which is a model that handles time-series information for combustion video data of various types of wood, to predict the combustion process of wood in order to …

AR 空間におけるデジタルコンテンツの評価支援を目的とした未就学児の行動解析

Authors

安藤廉, 宇佐美裕康, 鈴木裕利, 柊和佑

Journal

電子情報通信学会技術研究報告; 信学技報

Published Date

2022/1/20

抄録 (和) ICT 領域の目覚ましい進歩に伴い, AI, IoT, AR 等, これらを用いた ICT 教育の導入が急速に進められている.愛知県小牧市にあるこまきこども未来館という施設では, デジタルコンテンツによる体験学習を提供している. コンテンツは, 3 ヶ月毎に更新して改善を行っているが, 未就学児の反応の状況を開発担当者の目視による情報に基づいて行っているために, 全てのコンテンツに対する反応を把握できておらず, コンテンツの評価が十分ではないという問題がある. そこで本稿では, 未就学児の行動をヒートマップにより可視化を実現する手法を提案する.

Fundamental study on automatic analysis of handball game by score scene extraction

Authors

Keisuke Kachi,Hiroyasu Usami,Hiroaki Sawano,Yuri Suzuki

Journal

IEICE Technical Report; IEICE Tech. Rep.

Published Date

2022/1/20

(in English) Data analysis of sports images is an important task for improving strategy and the technical skills of athletes, but data analysis is often performed manually in minor sports, resulting in time and human costs.

ハンドボールの競技映像からの得点シーン抽出による分析自動化の基礎検討

Authors

可知奎介, 宇佐美裕康, 澤野弘明, 鈴木裕利

Journal

電子情報通信学会技術研究報告; 信学技報

Published Date

2022/1/20

抄録 (和) スポーツ映像のデータ分析は, 戦略面や選手の技術力向上を目的とした重要なタスクであるが, マイナースポーツにおいてはデータ分析を手動で行っている場合が多く, 時間と人的コストがかかってしまうという問題がある. スポーツ競技において得点に変化のあるシーンは特に重要であるが, 例えばハンドボール競技の場合では, 内容によっては試合映像が 2 時間にも及ぶ場合がある. そこで本稿では, 試合要約を目的として, 試合映像中に得点板が表示されているシーンを得点変化に関連するシーンと仮定し, 効果的にスコアボードが表示されているシーンを抽出する手法を提案する. 提案手法では, 背景差分法の応用による競技映像からの得点シーン抽出処理を行うことで, データ分析コスト低減を目指した. 結果として, シーン抽出の見逃しの評価指標である Recall にて 100% の結果を得ることができたので報告する.

Automatic generation of polyp image using depth map for endoscope dataset

Authors

Haruki Yamane,Shinji Fukui,Yuji Iwahori,Hiroyasu Usami,MK Bhuyan,Naotaka Ogasawara,Kunio Kasugai

Journal

Procedia Computer Science

Published Date

2021/9/8

In recent years, opportunities for diagnosis using endoscopy aiming a less invasive treatment are increasing following the disease rate of colorectal cancer. Computer-aided diagnosis has been developed based on deep learning methodology, it aiming to improve the accuracy of diagnosis and support immature medical doctors. To satisfy the learning dataset, this paper proposes a data augmentation methodology where automatic image generation of polyp images using Pix2Pix and depth map obtained from the original image. The problem of lack of the learning dataset of polyp images can be solved by the proposed approach and the effectiveness of the generated data was confirmed by the quantitative evaluation with the improved performance of SSD (Single Shot Multibox Detector) in the experiments.

PoC 貧乏を防ぐファクター X について-画像処理技術からの一考察

Authors

宇佐美裕康

Journal

IEICE Conferences Archives

Published Date

2021/2/23

近年の深層学習をはじめとするAI技術の大躍進により,画像処理技術の分野では,2015年には画像認識における人間の認識率をAIが超越し,実用化への期待が高まっている.AI技術の実用化に際しては,課題解決の手段としてAIによる実装が有効な手段かを一般的にPoC検証を行うが,PoCの段階で幾度となく検証を行うこととなり,PoC貧乏のような用語までうまれている.そのように,AI技術の実用化を阻む大きな壁が存在するのが現実である.これは,殆どの場合,コンセプト詳細を言語ベースで落とし込むことに起因し,非言語である質的情報が欠落してしまうためにPoC貧乏に陥る.本稿では,画像処理技術の視座からこれら課題の要となるファクターXについて考察を述べる.

Recovering shape from endoscope image using eikonal equation

Authors

Yuji Iwahori,Hiroyasu Usami,M. K. Bhuyan,Aili Wang,Naotaka Ogasawara,Kunio Kasugai

Published Date

2021/4

This paper proposed a shape recovery approach from Endoscope Image using Eikonal Equation. Photometric constraint equation derived from the Lambert reflectance and geometrical constraint equation derived from the relationship between the neighboring points are used and these equations can make a new approximation equation of Eikonal equation under the point light source illumination and perspective projection. The original endoscope image is transformed and generated to the Lambertian image by removing the specular reflectance. Framework of Fast Marching Method using the derived Eikonal Equation can recover the 3D shape from endoscope image. Usefulness was confirmed using simulation and experiments.

Diffusion patterns of social network posts

Authors

Alexander Gubanov,Yuliya Mundrievskaya,Ida M Pu

Published Date

2020

The “Vkontakte”[1] is one of the largest social networks on the previous-soviet space with 80-100 millions visitors daily. Nodes of the network represent users (individuals or groups), and ties can be a mutual (undirected) link, such as friendship or/and directed link, such as subscribing. Users exchange information by means of private messages or/and public posts. Posts are publications in texts or/and multimedia (images, sound and videos) on webpages. Recurrent posts are referred to as reposts of the original post, and the dedicated display areas for posts are referred to as walls. Figure 1 shows a screenshot as an example of the wall and users’ posts on the Vkontakte site.Posts on the wall are also queued in a newsfeed (poster) and become visible on devices of subscribed users (users for short hereafter). Any post can be reposted by other users, and appears on their walls and in newsfeeds of subscribers (as friends or/and followers). With such iterative processes, multiple chains of posts are formed and information is spread like epidemic.

Mediastinal Lymph Node Detection using Deep Learning.

Authors

Jayant P Singh,Yuji Iwahori,Manas Kamal Bhuyan,Hiroyasu Usami,Taihei Oshiro,Yasuhiro Shimizu

Published Date

2020/1

Accurate Lymph Node detection plays a significant role in tumour staging, choice of therapy, and in predicting the outcome of malignant diseases. Clinical examination to detect lymph node metastases alone is tedious and error-prone due to the low contrast of surrounding structures in Computed Tomography (CT) and to their varying shapes, poses, sizes, and sparsely distributed locations.(Oda et al., 2017) report 84.2% sensitivity at 9.1 false-positives per volume (FP/vol.) by local intensity structure analysis based on an Intensity Targeted Radial Structure Tensor (ITRST). In this paper, we first operate a candidate generation stage using U-Net (modified fully convolutional network for segmentation of biomedical images), towards 100% sensitivity at the cost of high FP levels to generate volumes of interest (VOI). Thereafter, we present an exhaustive analysis of approaches using different representations (ways to decompose a 3D VOI) as input to train Convolutional Neural Network (CNN), 3D CNN (convolutional neural network using 3D convolutions) classifier. We also evaluate SVMs trained on features extracted by the aforementioned CNN and 3D CNN. The candidate generation followed by false positive reduction to detect lymph nodes provides an alternative to compute and memory intensive methods using 3D fully convolutional networks. We validate approaches on a dataset of 90 CT volumes with 388 mediastinal lymph nodes published by (Roth et al., 2014). Our best approach achieves 84% sensitivity at 2.88 FP/vol. in the mediastinum of chest CT volumes.

See List of Professors in Hiroyasu Usami University(Chubu University)

Hiroyasu Usami FAQs

What is Hiroyasu Usami's h-index at Chubu University?

The h-index of Hiroyasu Usami has been 5 since 2020 and 6 in total.

What are Hiroyasu Usami's top articles?

The articles with the titles of

第 40 回ファジィシステムシンポジウム (FSS2024) 開催案内

Extraction of sample area in biomass ash sample dissolution data using time-series information

深層学習における簡素情報量を切り出す互助組織化メカニズムの確立

Blood Vessel Structure Analysis using a Simulation Model for the Purpose of Polyp Shape Recovery from Endoscopic Images

内視鏡画像からのポリープ形状復元を目的としたシミュレーションモデルによる血管構造解析

O-18 バイオマス灰の溶融試験法とその解釈に関する一考察

Seq2Seq を用いた木材に含まれる無機化合物の溶融過程の予測

Improvement of Polyp Detection using MUNIT for Image Generation

...

are the top articles of Hiroyasu Usami at Chubu University.

What are Hiroyasu Usami's research interests?

The research interests of Hiroyasu Usami are: computer vision, medical image processing, machine learning, graph theory

What is Hiroyasu Usami's total number of citations?

Hiroyasu Usami has 118 citations in total.

What are the co-authors of Hiroyasu Usami?

The co-authors of Hiroyasu Usami are Boonserm Kijsirikul, Yuji Iwahori.

    Co-Authors

    H-index: 21
    Boonserm Kijsirikul

    Boonserm Kijsirikul

    Chulalongkorn University

    H-index: 18
    Yuji Iwahori

    Yuji Iwahori

    Chubu University

    academic-engine

    Useful Links