Article
Computer Science, Artificial Intelligence
Xin Ding, Yongwei Wang, Zuheng Xu, William J. Welch, Z. Jane Wang
Summary: This article introduces the continuous conditional generative adversarial network (CcGAN), the first model designed for conditional generative modeling of image data with continuous, scalar conditions (known as regression labels). The article addresses the challenges posed by conditioning on regression labels, including the scarcity of real images for some labels and the inability to apply conventional label input mechanisms. The proposed CcGAN solves these problems by reformulating existing empirical cGAN losses, introducing new label input mechanisms, and proposing novel losses. Extensive experiments demonstrate that CcGAN outperforms cGAN both visually and quantitatively in generating high-quality samples from the image distribution.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Environmental Sciences
Qingli Luo, Hong Li, Zhiyuan Chen, Jian Li
Summary: Synthetic aperture radar (SAR) imagery provides all-day and all-weather observation, but its interpretation is difficult for nonexperts. Transferring SAR imagery into optical imagery can improve the interpretation of SAR data. The proposed ADD-UNet method based on cGAN shows superiority in SAR-to-optical translation.
Article
Chemistry, Multidisciplinary
Hiskias Dingeto, Juntae Kim
Summary: While Machine Learning has security flaws, this paper proposes a Universal Adversarial Training algorithm using an AC-GAN to generate adversarial examples. By enhancing the AC-GAN architecture and comparing its performance to other models, it is shown that generative models are better suited for boosting adversarial security through adversarial training.
APPLIED SCIENCES-BASEL
(2023)
Article
Computer Science, Artificial Intelligence
Yueqing Wang, Liang Deng, Yunbo Wan, Zhigong Yang, Wenxiang Yang, Cheng Chen, Dan Zhao, Fang Wang, Yang Guo
Summary: This study proposes an intelligent method based on a conditional generative adversarial network (cGAN) to predict the pressure coefficients (Cp) of airfoil. The method addresses existing issues and achieves a significant speedup compared with traditional computational fluid dynamics (CFD) simulation.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Biology
S. Jayalakshmy, Gnanou Florence Sudha
Summary: This study utilizes conditional generative adversarial networks (cGAN) for data augmentation to achieve higher accuracy in respiratory signal classification. The effectiveness of this method is verified through calculating similarity measures between signals and using various pre-trained deep learning architectures.
COMPUTERS IN BIOLOGY AND MEDICINE
(2021)
Article
Engineering, Electrical & Electronic
Zhirui Liang, Robert Mieth, Yury Dvorkin
Summary: This paper proposes a modified cGAN model to generate statistically credible net load scenarios for power systems, conditioned by given labels (e.g., seasons), that are stressful to system operations and dispatch decisions. The proposed OA-cGAN internalizes a DC optimal power flow model and seeks to maximize operating cost by generating worst-case data. The model is trained and tested using historical net load forecast errors.
ELECTRIC POWER SYSTEMS RESEARCH
(2022)
Article
Geochemistry & Geophysics
Zixu Wang, Shoudong Wang, Chen Zhou, Wanli Cheng
Summary: Neural networks are commonly used for seismic inversion, but their accuracy depends on the availability of labeled data. To overcome this, a closed-loop multitask conditional Wasserstein generative adversarial network (CMcWGAN) is proposed for AVO inversion. Experimental results show that CMcWGAN outperforms traditional methods in accuracy and robustness to noise.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Chemistry, Multidisciplinary
Kyungho Yu, Juhyeon Noh, Hee-Deok Yang
Summary: This paper presents a method for automatically extracting line drawings representing geometric characteristics from cartoon images, aiming to shorten the time required for creating 3D models. Experimental results demonstrate the effectiveness of the proposed method.
APPLIED SCIENCES-BASEL
(2021)
Article
Engineering, Electrical & Electronic
Zeynel Deprem, A. Enis Cetin
Summary: The article discusses the value of signal representation in the time-frequency (TF) domain and its applications in various fields. A novel method using conditional generative adversarial network (CGAN) for signal reassignment is proposed, which generates high-resolution TF representations better than current methods.
SIGNAL IMAGE AND VIDEO PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Yi Sima, Jizheng Yi, Aibin Chen, Ze Jin
Summary: Using two key-frames with neutral expression and maximum expression intensity to alleviate inter-subject variations in Facial Expression Recognition (FER). Generating key-frames using conditional Generative Adversarial Network and triplet distance model, designing two-stream CNN model for extracting differential emotion features. Extensive comparisons on different databases demonstrating the superiority of the proposed GDCNN framework.
APPLIED SOFT COMPUTING
(2021)
Article
Neurosciences
Chong Wang, Hongmei Yan, Wei Huang, Jiyi Li, Yuting Wang, Yun-Shuang Fan, Wei Sheng, Tao Liu, Rong Li, Huafu Chen
Summary: Recent fMRI studies have made progress in reconstructing visual content, but reconstructing dynamic natural vision is still challenging. In this study, a novel fMRI-conditional video generative adversarial network was developed to reconstruct rapid video stimuli from fMRI responses, capturing important spatiotemporal information.
Article
Computer Science, Artificial Intelligence
Xin Ding, Yongwei Wang, Zuheng Xu, Z. Jane Wang, William J. Welch
Summary: Knowledge distillation is actively studied in deep learning for image classification tasks. This paper proposes a comprehensive knowledge distillation framework based on conditional generative adversarial networks, which is suitable for both classification and regression tasks, and is flexible and insensitive to the architecture of teacher and student models.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Chemistry, Analytical
Minjae Kang, Yong Seok Heo
Summary: In this paper, a new model called GammaGAN is proposed for conditional video generation. The model effectively utilizes class labels through the projection method and introduces scaling class embeddings and normalizing outputs, leading to improved video quality. Evaluation results demonstrate relative improvements in PSNR, SSIM, and LPIPS compared to a prior model, indicating potential for further advancements in conditional video generation.
Article
Computer Science, Artificial Intelligence
Xin Ding, Yongwei Wang, Z. Jane Wang, William J. Welch
Summary: To address the inefficiency in subsampling images generated from conditional generative adversarial networks (cGANs), we propose a novel approach called conditional density ratio rejection sampling (cDR-RS). It incorporates an improved feature extraction mechanism and a conditional Softplus loss (cSP) to estimate the conditional density ratio, allowing for the acceptance or rejection of fake images. Additionally, a filtering scheme is introduced to enhance label consistency without compromising diversity when sampling from continuous cGANs (CcGANs). Experimental results show that cDR-RS achieves state-of-the-art performance in subsampling both class-conditional GANs and CcGANs.
Article
Computer Science, Interdisciplinary Applications
Francis Baek, Daeho Kim, Somin Park, Hyoungkwan Kim, SangHyun Lee
Summary: This study proposes a data augmentation method that combines generative adversarial networks with a target classifier to address the issue of data shortage in computer vision applications for construction. The results demonstrate that the proposed method significantly improves classification accuracy and enhances feature extraction for the target object.
JOURNAL OF COMPUTING IN CIVIL ENGINEERING
(2022)