Papers - YAMAMORI Kunihito
-
Implementation of a highly adaptive interior simulator using GrabCut and LAMA
Seishin Yamamura, Kunihito Yamamori
A-3-1 2025.3
Authorship：Corresponding author Language：Japanese Publishing type：Research paper (conference, symposium, etc.)
-
Interior layout image generation by the combination of GrabCut and Large-mask inpainting algorithm with 3D furniture models Reviewed
Seishin Yamamura, Kunihito Yamamori, Ryo SAITO
Proceedings of the Joint Symposium of The Thirtieth International Symposium on Artificial Life and Robotics (AROB 30th 2025) 1101 - 1106 2025.1
Authorship：Corresponding author Language：English Publishing type：Research paper (international conference proceedings)
-
Constaria: a chatbot for long-term consistent conversation Reviewed
Shuta Yamane, Daisuke Amaki, Kunihito Yamamori
Proceedings of the Joint Symposium of The Thirtieth International Symposium on Artificial Life and Robotics (AROB 30th 2025) 1117 - 1122 2025.1
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings)
-
An active listening and responding system with empathy enhancing filter using ChatGPT Reviewed
Daisuke AMAKI, Shuta YAMANE, Kunihito YAMAMORI
Proceedings of the Joint Symposium of The Thirtieth International Symposium on Artificial Life and Robotics (AROB 30th 2025) 1101 - 1106 2025.1
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings)
-
An approach for driving manner improvement by a gamepad-controlled driving simulator Reviewed
Kotaro SHIIKI, Shuta YAMANE, Kunihito YAMAMORI
Proceedings of the Joint Symposium of The Thirtieth International Symposium on Artificial Life and Robotics (AROB 30th 2025) 1107 - 1111 2025.1
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings)
-
Transformer-based video captioning with object motion in video
Koya Kimura, Ryo Saito, Kunihito Yamamori
2024.3
Authorship：Corresponding author Language：Japanese Publishing type：Research paper (conference, symposium, etc.)
-
Twin-Encoder Approach for Automatic Video Captioning Reviewed
Koya Kimura, Ryo Saito, Kunihito Yamamori
Proceedings of 29th International Symposium on Artificial Life and Robotics 981 - 922 2024.1
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings)
-
Robustness comparison of machine learning algorithms for NIDS under the same environment Reviewed
Masaki Tagawa, Kunihito Yamamori, Masaru Aikawa
Proceedings of the Joint Symposium of The Twenty-Eighth International Symposium on Artificial Life and Robotics (AROB 28th 2023) 931 - 935 2023.1
Authorship：Corresponding author Language：English Publishing type：Research paper (international conference proceedings)
-
Performance of Machine Learning base NIDS on Re-organized Kyoto 2016 Dataset Reviewed
Ryo Saito, Masaru Aikawa, Kunihito Yamamori
Proceedings of the Joint Symposium of The Twenty-Eighth International Symposium on Artificial Life and Robotics (AROB 28th 2023) 926 - 930 2023.1
Authorship：Corresponding author Language：English Publishing type：Research paper (international conference proceedings)
-
Heuristic base music arrangement suppressing on discord progression Reviewed
Kosuke Yoshida, Masaru Aikawa, Kunihito Yamamori
Proceedings of the Twenty-Seventh International Symposium on Artificial Life and Robotics 2022 1052 - 1056 2022.1
Authorship：Corresponding author Language：English Publishing type：Research paper (international conference proceedings)
-
Driving trajectory optimization by reinforcement learning for motorsports Reviewed
Akinobu Iwai, Masaru Aikawa, Kunihito Yamamori
Proceedings of the Twenty-Seventh International Symposium on Artificial Life and Robotics 2022 1047 - 1051 2022.1
Authorship：Corresponding author Language：English Publishing type：Research paper (international conference proceedings)
-
Study on Email Spam Classification Using Machine Learning Techniques International coauthorship
Thae Ma Ma, Kunihito YAMAMORI, Masaru AIKAWA, Aye Thida
Memoirs of Faculty of Engineering, Miyazaki University 50 113 - 118 2021.9
Authorship：Corresponding author Language：English Publishing type：Research paper (bulletin of university, research institution)
-
Network Design for Session Type NIDS Reviewed
Ryo Saito, Kunihito Yamamori, Masaru Aikawa, Kentaro Inoue
PROCEEDINGS OF THE TWENTY-SIXTH INTERNATIONAL SYMPOSIUM ON ARTIFICIAL LIFE AND ROBOTICS 518 - 523 2021.1
Language：English Publishing type：Research paper (international conference proceedings)
In this paper, we propose a network design for session-type Network-based Intrusion Detection System (NIDS).
We focus on the network itself, network availability and stream data handling to achieve the practical session type NIDS. NIDS insertion cost depends on the network design, and it tends to increase. When the inserted NIDS fails, communications may be blocked depending on the insertion style. We use the Address Resolution Protocol (ARP) to reduce the cost of inserting NIDS into the network. And, our design prepares an independent controller from NIDS to improve the network vailability. After inserting the NIDS into the network, NIDS starts monitoring of the packets. And a large amount of memory is required to build a session from the packets. The proposed NIDS only focuses on the headers of some protocols to reduce memory usage. -
Tuning Support Tool forWAF Mod Security by Log Analysis with Machine Learning Reviewed
Chihiro Kudo, Kunihito Yamamori, Masaru Aikawa, Kentaro Inoue, Ryo Saito
PROCEEDINGS OF THE TWENTY-SIXTH INTERNATIONAL SYMPOSIUM ON ARTIFICIAL LIFE AND ROBOTICS 539 - 543 2021.1
Language：English Publishing type：Research paper (international conference proceedings)
In this research, we develop and evaluate a tool that supports tuning work of WAF (Web Application Firewall). The tuning of WAF is very hard because a lot of new vulnerabilities are found in every day, and deep understanding of computer and network securities are required. At the tuning work, attack detection rules are set to reduce misdetections. Misdetection has two attributes, false positive and false negative. In this research, we aim to reduce false positives because WAF operators has to deal with every positive detection for secure operation of their web services. Our tool makes a classifier train by machine learning algorithm with a dataset including true positives and false positives, and the classifier obtains rules how to classify a
communication is an attack or not. When a false positive detection will occur, then our tool gives the WAF operators an advice to correct the attack detection rules. -
Reinforcement Learning Approach for Imperfect Information Games Reviewed
Akinobu Iwai, Kunihito Yamamori, Masaru Aikawa, Kentaro Inoue
PROCEEDINGS OF THE TWENTY-SIXTH INTERNATIONAL SYMPOSIUM ON ARTIFICIAL LIFE AND ROBOTICS 533 - 538 2021.1
Language：English Publishing type：Research paper (international conference proceedings)
In this paper, we show whether Reinforcement Learning(RL) is effective or not for incomplete information game. We apply RL approach to a multiplayer incomplete information game ”Inferno”. RL is one of the machine learning algorithm to learn policies to choose an action of agent at the specific situation in virtual simulation environment. RL agents learn what action choices to make to win the game. RL plays the game against two different bots. We evaluate the proposed method by the score.
-
Computer aided arrangements using cadences and melodic regularity Reviewed
Kosuke Yoshida, Kunihito Yamamori, Masaru Aikawa, Kentaro Inoue
PROCEEDINGS OF THE TWENTY-SIXTH INTERNATIONAL SYMPOSIUM ON ARTIFICIAL LIFE AND ROBOTICS 529 - 532 2021.1
Language：English Publishing type：Research paper (international conference proceedings)
In this paper, we aim to propose and evaluate music arrangement with computer support. We use cadences and
melodic regularity in music arrangement. Cadences are musical heuristics and rules of transition easiness from a chord to the other chord. Since a lot of music has a regularity that includes the repetition of the same melody within a song, we use regularity to add a chord progression. To evaluate our proposed method, we get our information from questionnaires. In this questionnaire,
naturalness of the music created by our method compared with the music created by cadence only. -
Delivery Route Suboptimization Combined by Map Application and Genetic Algorithm Reviewed
Yoshitaka Matsushita, Masaru Aikawa, Kentaro Inoue, Kunihito Yamamori
PROCEEDINGS OF THE TWENTY-SIXTH INTERNATIONAL SYMPOSIUM ON ARTIFICIAL LIFE AND ROBOTICS 524 - 528 2021.1
Language：English Publishing type：Research paper (international conference proceedings)
In this paper, we aim to propose and evaluate a delivery route optimization method including random redelivery
requests. An arrangement of Genetic Algorithm and enumeration method is used to determine the delivery route. Genetic algorithm and enumeration method are one of the popular solution methods to solve Traveling Salesman Problem. During the experiments, a map application will be used to obtain real distances. In the environment for experiments, absent destinations are
randomly arranged and redelivery requests also randomly occur. Experiments showed that our method could present the delivery route even if 30% of destinations were absent and 60% of those sent redelivery requests when we set the number of destinations to 20. -
Ma T.M., Yamamori K., Thida A.
2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020 324 - 326 2020.10
Authorship：Corresponding author Language：Japanese Publishing type：Research paper (scientific journal) Publisher：2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020
Spam or unsolicited emails that are used by spammers can cause huge loss to both the email users and the email server. Therefore, in order to detect spam emails not to enter into our mailbox, a developed email spam classification system is required. This paper proposes two popular machine learning methods, Naïve Bayes Classifier and Support Vector Machine, to classify the emails into spam or ham based on the body or content of the emails. In Naïve Bayes Classifier, independent words are considered as features. Support Vector Machine can be used to represent an email in vector space in which each feature means one dimension. Finally, two methods are compared in terms of precision, recall, F-measure performance metrics with the aim of finding the best method.
-
Fast Retinex Image Enhancement Using CUDA International coauthorship
Kyi Myo Zaw, Kunihito Yamamori, Masaru Aikawa, Aye Min Myat
Memoirs of Faculty of Engineering, Miyazaki University 49 239 - 244 2020.9
Authorship：Corresponding author Language：English Publishing type：Research paper (bulletin of university, research institution)
-
Affect of data unbalance in "Kyoto 2016 Dataset" for NIDS with machine learning Reviewed
Ryo Saito, Masaru Aikawa, Kentaro Inoue, Kunihito Yamamori
Proceedings of The Twenty-fifth International Symposium on Artificial Life and Robotics 612 - 616 2020.1
Language：English Publishing type：Research paper (international conference proceedings)