Article
Computer Science, Artificial Intelligence
Lu Yi, Man-Wai Mak
Summary: This article proposes an adversarial data augmentation network based on generative adversarial networks (GANs) to address the overfitting problem in training deep neural networks with scarce data. The proposed networks generate augmented data rich in emotion information and yield competitive emotion classifiers.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Baptiste Roziere, Morgane Riviere, Olivier Teytaud, Jeremy Rapin, Yann LeCun, Camille Couprie
Summary: The study proposes a simple strategy to learn new generations of images from a dataset chosen by the user, while providing some control over the generation process. By recovering optimal latent parameters from the model through optimization steps, the generated images can better reflect the user's preferences.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Software Engineering
Jianan Li, Jimei Yang, Jianming Zhang, Chang Liu, Christina Wang, Tingfa Xu
Summary: The article introduces a new method, Attribute-conditioned Layout GAN, for generating graphic layouts constrained by design element attributes. By incorporating element dropout method and loss designs based on different design principles, the layout generation process is optimized. The effectiveness of the method is validated through a user study.
IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS
(2021)
Article
Engineering, Biomedical
Dongdong Li, Zhuo Yang, Zhe Wang, Ming Hua
Summary: We propose a framework that combines generative adversarial networks and speaker recognition to generate speaker-related emotional training speech features, enhancing robustness under different emotional conditions. The framework includes a new speaker emotion-converted generative adversarial network (SEC-GAN) for speaker recognition. SEC-GAN learns speech information from neutral speech and generates speech features in other emotions while retaining speaker identity. By using the Mandarin Affective Speech Corpus (MASC), our framework reduces the negative impact of emotion mismatch between speech. This strategy solves the common problem of voice control devices failing to recognize user identity when users speak in different emotions.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2023)
Article
Computer Science, Information Systems
Hajar Emami, Majid Moradi Aliabadi, Ming Dong, Ratna Babu Chinnam
Summary: This paper introduces a novel SPA-GAN model with attention mechanism for improved image-to-image translation tasks, as well as an additional feature map loss during training to preserve domain specific features. Compared to existing attention-guided GAN models, SPA-GAN demonstrates superior performance without the need for additional attention networks or supervision.
IEEE TRANSACTIONS ON MULTIMEDIA
(2021)
Article
Geochemistry & Geophysics
Tanmoy Dam, Sreenatha G. Anavatti, Hussein A. Abbass
Summary: A three-player spectral generative adversarial network architecture was proposed to handle minority classes under imbalanced conditions, using a class-dependent mixture generator spectral GAN. The method was validated in experiments showing effectiveness in generating minority class data and improving performance.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Computer Science, Information Systems
Shabab Bazrafkan, Viktor Varkarakis, Joseph Lemley, Hossein Javidnia, Peter Corcoran
Summary: The study introduces a new framework for training deep conditional generators by placing a classifier or regression model in parallel with the discriminator and back-propagating classification or regression errors. Experimental results on multiple datasets and mathematical proofs show that the method can achieve better results compared to other techniques.
Article
Computer Science, Artificial Intelligence
Yoon-Jae Yeo, Yong-Goo Shin, Seung Park, Sung-Jea Ko
Summary: This brief presents a simple but effective method to improve the performance of generative adversarial networks (GANs) by introducing a cascading rejection module for the discriminator.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Tao Hu, Chengjiang Long, Chunxia Xiao
Summary: This study focuses on utilizing images and text descriptions from social media platforms to build large-scale labeled datasets, proposing a novel visual text representation method that uses synthetic images generated by diverse conditional Generative Adversarial Network for visual recognition. By combining visual features extracted from synthetic images with text features, the effectiveness of the proposed representation for visual recognition is demonstrated through extensive experiments on benchmark datasets.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Information Systems
Aiwen Jia, Zhen-Hong Jia, Jie Yang, Nikola K. Kasabov
Summary: The paper introduces a single-image snow removal method based on a generative adversarial network and attention mechanism, which can better handle snow-covered areas by combining U-Net and residual network to enhance model performance. Experimental results show that this method produces better results than other state-of-the-art methods.
Article
Computer Science, Information Systems
Xiaopeng Chao, Jiangzhong Cao, Yuqin Lu, Qingyun Dai, Shangsong Liang
Summary: Generative Adversarial Networks (GANs) are a powerful subclass of generative models, but effectively training them to reach Nash equilibrium is a challenge. One possible solution is to bound the function space of the discriminator, which can help to reach Nash equilibrium faster and obtain better generative data during training.
Article
Engineering, Electrical & Electronic
Yifan Xia, Wenbo Zheng, Yiming Wang, Hui Yu, Junyu Dong, Fei-Yue Wang
Summary: Facial expression synthesis has gained attention with the development of GANs. The proposed LGP-GAN method extracts and synthesizes the details of crucial facial regions using a two-stage cascaded structure, resulting in superior performance compared to state-of-the-art methods.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2022)
Article
Automation & Control Systems
Huaping Liu, Di Guo, Xinyu Zhang, Wenlin Zhu, Bin Fang, Fuchun Sun
Summary: This article introduces a portable device that provides tactile recognition assistance for visually impaired people, with potential extensive applications in tactile mouse and white cane. The developed technology can be widely utilized in various industrial applications such as surrounding monitoring and manipulation. The proposed work demonstrates the promising ability of artificial intelligence in healthcare applications, and the generated tactile signals are expected to be used in many human-centered systems, marking an important step towards the development of a more comprehensive assisting technology for visually impaired people.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2021)
Article
Computer Science, Artificial Intelligence
Xiaoning Song, Yao Chen, Zhen-Hua Feng, Guosheng Hu, Dong-Jun Yu, Xiao-Jun Wu
Summary: The article presents a new Self-growing and Pruning Generative Adversarial Network (SP-GAN) for realistic image generation, which dynamically adjusts the network size and architecture, and utilizes an adaptive loss function. Experimental results show the method's advantages in stability and efficiency.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2021)
Article
Computer Science, Information Systems
Le Minh Ngo, Sezer Karaoglu, Theo Gevers
Summary: A novel architecture is proposed in this paper for manipulating facial expressions, head poses, and lighting conditions from a single monocular image. The method outperforms state-of-the-art methods in various scenarios and does not require target specific training.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)