IEICE Technical Report

Online edition: ISSN 2432-6380

Pattern Recognition and Media Understanding

Workshop Date : 2024-11-29 - 2024-11-30 / Issue Date : 2024-11-22

PRMU2024-8
Task Partitioning and Ordering for Continual Learning -- Class Clustering from Confusion Matrix --
Toshikazu Wada (Wakayama Univ.)
pp. 1 - 6

PRMU2024-9
Performance Evaluation of DINO with Attention Modification Mechanism Introduced by Sigmoid Attention
Yuki Matsumoto (SEI), Tomokazu Ozawa (Unico), Takuya Okuno (SCE), Katsushi Miura (SEI)
pp. 7 - 12

PRMU2024-10
Learning Neural Velocity Fields using Edge-based Ray Sampling
Sota Ito, Yoshikazu Hayashi (Gifu Univ.), Hiroaki Aizawa (Hiroshima Univ), Kunihito Kato (Gifu Univ.)
pp. 13 - 18

PRMU2024-11
Self-supervised Contrastive Learning for Remote Sensing Imagery with Perlin noise injection
Riku Okazawa, Eisaku Maeda (Tokyo Denki Univ.)
pp. 19 - 24

PRMU2024-12
[Tutorial Lecture] Real-world, 3D, Robotics and Language Understanding -- Beyond vision and language understanding --
Shuhei Kurita (NII)
p. 25

PRMU2024-13
Theoretical Explanation for Learning from Adversarial Perturbations
Soichiro Kumano (UTokyo), Hiroshi Kera (Chiba Univ.), Toshihiko Yamasaki (UTokyo)
pp. 26 - 31

PRMU2024-14
(See Japanese page.)
pp. 32 - 37

PRMU2024-15
Concept acquisition in the CNN learning process
Shun Iwase, Reiya Hiramoto, Keito Kobayashi, Eisaku Maeda (TDU)
pp. 38 - 43

PRMU2024-16
Robustness of DNN Models against Model Extraction Attacks
Reiya Hiramoto, Shun Iwase, Eisaku Maeda (TDU)
pp. 44 - 49

PRMU2024-17
Determination of cardiac stenosis patients by CNN using 3D non-contrast cardiac CT data
Tetsuya Asakawa (TUT), Masashi Hashimoto, Takeshi Miyaji, Kei Nomura, , Kazuki Shimizu (THC), Masaki Aono (TUT)
pp. 50 - 54

PRMU2024-18
Bezier Curve-Based Stroke Extraction for Handwritten Characters
Yuxuan Teng, Takuya Matsuzaki (TUS)
pp. 55 - 60

PRMU2024-19
Proposal of Weakly Supervised Disentanglement and Its Application to Text Image Analysis
Hayato Mitani (Kyushu Univ.), Daichi Haraguchi, Wataru Shimoda, Naoto Inoue, Kota Yamaguchi (CyberAgent), Seiichi Uchida (Kyushu Univ.)
pp. 61 - 66

PRMU2024-20
Composition of OCR Training Data towards Automated Marking of Transcription
Ayumi Seki, Takuya Matsuzaki (TUS)
pp. 67 - 71

PRMU2024-21
Completion of Mokkan Character Images by Diffusion Model
Kenji Shiono, Wataru Ohyama (Tokyo Denki Univ.)
pp. 72 - 76

PRMU2024-22
Popularity Prediction on Fashion SNS with Multimodal Features
Shun Okamoto, Satoshi Iizuka, Kazuhiro Fukui (Tsukuba Univ.)
pp. 77 - 82

PRMU2024-23
A Study on Mental Health Estimation from Dietary Photo Logs Using Large Multimodal Models
Yuki Imajuku, Yoko Yamakata, Kiyoharu Aizawa (UTokyo)
pp. 83 - 88

PRMU2024-24
Matsue city rivers flood forecasting method based on gaussian process autoregression
Kaisho Wakabayashi, Hitoshi Sakano, Hiroshi Yajima (Shimane University)
pp. 89 - 92

PRMU2024-25
Disentangle Learning for Interactive Semiconductor Inspection Robust to Thin-Film Interference
Koki Mitsusada, Koji Kamma, Toshikazu Wada (Wakayama Univ.), Yasuhiro Hamada, Yoshinori Shimada, Takeshi Noguchi, Toshiyuki Okayama (Advanced System Solutions Co., Ltd.)
pp. 93 - 98

PRMU2024-26
[Invited Talk] Deep image synthesis with physical and geometrical constraints
Takuhiro Kaneko (NTT)
p. 99

PRMU2024-27
[Invited Talk] Graphic Design Generation with Multi-modal Models
Naoto Inoue (CA)
p. 100

PRMU2024-28
Feature Analysis of Adversarial Examples in Perception and Reconstruction
Takumi Maeda (Koshi High School), Soichiro Kumano, Toshihiko Yamasaki (UTokyo)
pp. 101 - 104

PRMU2024-29
(See Japanese page.)
pp. 105 - 110

PRMU2024-30
A New Weighting Strategy for Group Robust Classification
Yuya Saito, Shinnosuke Matsuo, Seiichi Uchida (Kyushu Univ.), Daiki Suehiro (Kyushu Univ./RIKEN AIP)
pp. 111 - 115

PRMU2024-31
Subspace-Based Decision Tree Retraining
Yoriaki Murata, Hitoshi Sakano (Shimane Univ)
pp. 116 - 119

PRMU2024-32
CommonArt -- Diffusion Transformer for Text-to-Image Generation by Japanese Large Language Model --
Yasunori Ozaki, Ryuji Mishima, Toshiki Tomihira
pp. 120 - 125

Note: Each article is a technical report without peer review, and its polished version will be published elsewhere.

The Institute of Electronics, Information and Communication Engineers (IEICE), Japan