Search

Your search keyword '"Shoma A."' showing total 181 results
Start Over Author "Shoma A." Publisher ieee
181 results on '"Shoma A."'

1. Dual-Conditioned Training to Exploit Pre-Trained Codebook-Based Generative Model in Image Compression

Author

Shoma Iwai, Tomo Miyazaki, and Shinichiro Omachi

Subjects
Generative adversarial networks, image compression, VQGAN, Electrical engineering. Electronics. Nuclear engineering, TK1-9971
Abstract

Learned image compression (LIC) is increasingly gaining attention. To improve the perceptual quality of reconstructions, generative LIC has been studied, using generative models such as Generative Adversarial Networks (GANs). State-of-the-art generative LIC methods have achieved remarkable performance even in low bit rate settings. Unlike most approaches trained from scratch, we propose a generative LIC that utilizes a pre-trained codebook-based generative model, Vector-Quantized GAN (VQGAN). Specifically, our model is designed to exploit its powerful image-generation capabilities to enhance compression performance. Our approach reconstructs an image from a transmitted bitstream in two steps: (1) estimating VQGAN tokens and feeding them into the pre-trained VQGAN decoder, and (2) modifying the decoder’s intermediate features to address artifacts and distortions. Our preliminary experiments reveal that the information allocation between (1) and (2) is pivotal for reconstruction quality. Moreover, we found that the ideal allocation varies based on the target bit rate. Motivated by these findings, we propose a novel Dual-Conditioned training. Through the training, the model learns to adjust the total bit rate and information allocation between (1) and (2) based on two conditional inputs. Subsequently, we explore the conditional inputs to achieve the optimal results for each target bit rate. This training strategy enables us to effectively exploit the generation capability of VQGAN across different bit rates. Our method, named Dual Conditioned VQGAN-based Image Compression (DC-VIC), outperforms state-of-the-art generative LIC methods in rate-distortion-perception performance. Code will be available at https://github.com/iwa-shi/DC_VIC

Published
2024
Full Text
View/download PDF

2. Semantically-Guided Image Compression for Enhanced Perceptual Quality at Extremely Low Bitrates

Author

Shoma Iwai, Tomo Miyazaki, and Shinichiro Omachi

Subjects
Image compression, semantic information, perceptual image compression, GANs, neural image compression, Electrical engineering. Electronics. Nuclear engineering, TK1-9971
Abstract

Image compression methods based on machine learning have achieved high rate-distortion performance. However, the reconstructions they produce suffer from blurring at extremely low bitrates (below 0.1 bpp), resulting in low perceptual quality. Although some methods attempt to reconstruct sharp images using Generative Adversarial Networks (GANs), reconstructing natural textures at low bitrates remains challenging. In this paper, we propose a novel image compression method that explicitly utilizes semantic information. Specifically, we send a semantic label map to the decoder, which takes it as input. This semantic information enables the decoder to reconstruct appropriate textures consistent with the corresponding semantic classes. Although semantic label maps can be compressed into relatively small data sizes using common methods (e.g., PNG), the data size is not negligible in an extremely low-rate setting. To address this problem, we propose simple yet effective label map compression strategies, including an autoregressive label map compressor. Our strategies significantly reduce the data size of the label map while maintaining the critical semantic information that allows the decoder to reconstruct realistic and suitable textures. By utilizing this data-efficient semantic information, our method can reconstruct realistic images even at an extremely low bitrate. As a result, the proposed method outperformed existing models, including a GAN-based model designed for low-rate settings and a state-of-the-art semantically guided method, in both quantitative evaluation and user studies. Furthermore, we analyzed the effect of semantic information by switching the input label map, confirming that the model synthesized textures appropriate to the given semantic labels.

Published
2024
Full Text
View/download PDF

3. Operation Range of Intermittent Velocity Control for Improving Top-Dead-Center Accuracy in Dual-Sided Free-Piston Engine Linear Generator

Author

Mitsuhide Sato, Shoma Irie, Tsutomu Mizuno, Yuhei Sakane, and Kaname Naganuma

Subjects
Free-piston engine, linear generator, generation control, speed control, and top-dead-center, Electrical engineering. Electronics. Nuclear engineering, TK1-9971
Abstract

A free-piston engine linear generator (FPEG) is constructed by combining a reciprocating internal combustion engine and a linear generator. FPEG is required to improve both the accuracy of the top-dead-center (TDC) positioning considering frequent combustion fluctuations and the reduction of power consumption. This paper confirms the influence of intermittent velocity control on power running suppression and TDC position deviation in a dual-cylinder FPEG. Intermittent velocity control is a simple method, which consists of repeatedly turning the speed control on and off. The experimental results of the glow engine indicate that this control suppresses power consumption and enables continuous combustion. The setting range for intermittent speed control, which continues operation only through regenerative operation, must be determined by considering the gain value. In addition, this paper explains the effects of the control range of intermittent velocity control on power generation efficiency, piston motion frequency, and TDC position deviation in the dual-cylinder FPEG. Increasing the range ratio of intermittent speed control improves power generation efficiency, but it also increases the standard deviation of TDC position.

Published
2023
Full Text
View/download PDF

30. Neural Architecture Search for Improving Latency-Accuracy Trade-off in Split Computing

Author

Shimizu, Shoma, Nishio, Takayuki, Saito, Shota, Hirose, Yoichi, Yen-Hsiu, Chen, and Shirakawa, Shinichi

Subjects
FOS: Computer and information sciences, Computer Science - Machine Learning, Computer Science - Distributed, Parallel, and Cluster Computing, Distributed, Parallel, and Cluster Computing (cs.DC), Machine Learning (cs.LG)
Abstract

This paper proposes a neural architecture search (NAS) method for split computing. Split computing is an emerging machine-learning inference technique that addresses the privacy and latency challenges of deploying deep learning in IoT systems. In split computing, neural network models are separated and cooperatively processed using edge servers and IoT devices via networks. Thus, the architecture of the neural network model significantly impacts the communication payload size, model accuracy, and computational load. In this paper, we address the challenge of optimizing neural network architecture for split computing. To this end, we proposed NASC, which jointly explores optimal model architecture and a split point to achieve higher accuracy while meeting latency requirements (i.e., smaller total latency of computation and communication than a certain threshold). NASC employs a one-shot NAS that does not require repeating model training for a computationally efficient architecture search. Our performance evaluation using hardware (HW)-NAS-Bench of benchmark data demonstrates that the proposed NASC can improve the ``communication latency and model accuracy" trade-off, i.e., reduce the latency by approximately 40-60% from the baseline, with slight accuracy degradation.

Published
2022
Full Text
View/download PDF

44. A Global Shutter Wide Dynamic Range Soft X-Ray CMOS Image Sensor With Backside- Illuminated Pinned Photodiode, Two-Stage Lateral Overflow Integration Capacitor, and Voltage Domain Memory Bank

Author

Shike, Hiroya, Kuroda, Rihito, Kobayashi, Ryota, Murata, Maasa, Fujihara, Yasuyuki, Suzuki, Manabu, Harada, Shoma, Shibaguchi, Taku, Kuriyama, Naoya, Hatsui, Takaki, Jun, Miyawaki, Harada, Tetsuo, Yamasaki, Yuichi, Watanabe, Takeo, Harada, Yoshihisa, and Sugawa, Shigetoshi

Abstract

This article presents a prototype 22.4 μm pixel pitch global shutter (GS) wide dynamic range (WDR) soft X-ray CMOS image sensor (sxCMOS). Backside-illuminated (BSI) pinned photodiodes with a 45-μm thick Si substrate were introduced for low noise and high radiation hardness to high energy photons. Two-stage lateral overflow integration capacitor (LOFIC) and voltage domain memory bankwith high-densitySi trench capacitorswere introduced for WDR and for GS. The developed sxCMOS achieved maximum 21.9 Me− full well capacity with a single exposure 129 dB dynamic range by GS operation. Over 70% quantum efficiency (QE) toward soft X-ray was successfully achieved. The developed prototype sxCMOS is a step forward toward a 4 M pixel detector system to be utilized in next-generation synchrotron radiation facilities and X-ray free-electron lasers.

Published
2021

50. Periodic Networked Imaging With Nanoscale Sensor Nodes via Two-Layered Time-Division Access

Author

Nishibori, Shoma, Murase, Tutomu, Tadokoro, Yukihiro, Nishibori, Shoma, Murase, Tutomu, and Tadokoro, Yukihiro

Abstract

As the Internet of Things (IoT) has become a widespread phenomenon, promising sensor applications at the nanoscale have begun to emerge. One example is imaging via distributed massive nanoscale nodes (NSNs), which can be used to implement “invisible surveillance cameras” by, for example, painting liquids containing nanoscale sensors onto walls. This imaging method requires periodic data transfer from thousands of NSNs to a data collection node (DCN). An essential technique for handling such transfers is the media access control (MAC) protocol. However, existing protocols cannot support periodic transfer from numerous NSNs because of inefficient communication caused by a large amount of headers in the packets. In this article, we introduce an original MAC protocol and discuss its capability in terms of implementing imaging applications. The proposed protocol applies a time-division access (TDA) feature to reduce the amount of headers and a two-layered protocol to enable simultaneous transmission among nodes. Slot assignment is an essential function in TDA and requires communication among nodes. Unlike existing methods, our simple approach enables communication by exploiting the unique features of the focused application. The results of the numerical simulation reveal that the proposed MAC protocol allows for periodic imaging with more than three thousand nodes and produces high-quality images very close to those obtained using ideal communication. These results are achieved by employing an original design framework to determine appropriate key parameters, such as the number of clusters and frame rate.

Published
2022

Catalog

Books, media, physical & digital resources
See catalog results