Application and research of dcgan model in image generation tasks
Lubin Liu
Shenzhen University School of Electronics and Information Engineering
DOI: https://doi.org/10.59429/esta.v11i4.8459
Keywords: Generative learning; DCGAN; Cartoon image generation; Handwritten digit generation; Conditional Generative Adversarial Network
Abstract
This article aims to delve into the principles of generative learning and its applications in image generation tasks by constructing cartoon image generation models and handwritten digit generation models. The experiment is based on the innovation of model architecture using Deep Convolutional Generative Adversarial Network (DCGAN), including adding convolutional layers to enlarge image size and applying Gaussian filtering and average pooling in cartoon image generation tasks, as well as designing conditional generative adversarial networks to generate handwritten images of specific numbers based on labels in handwritten digit generation tasks. The experiment elaborated on the model architecture, data processing, training process, and testing methods, and explored the impact of different network architectures and parameter settings on model performance. The results indicate that the proposed model architecture can effectively improve the quality of image generation, providing new ideas for the research and application of generative learning.
References
[1] Xie Tianqi, Wu Yuanyuan, Jing Chao, Sun Weiheng Overview of GAN Model Generated Image Detection Methods [J]. Computer Engineering and Applications, 2024, 60 (22): 74-86.
[2] Zhao Hong, Li Wengai Research on Text Generation Image Model Based on Diffusion Generative Adversarial Network [J]. Chinese Journal of Electronics and Information Technology, 2023, 45 (12): 4371-4381.
[3] Liu Zerun, Yin Yufei, Xue Wenhao, Guo Rui, Cheng Lechao A review of conditional guided image generation based on diffusion models [J]. Journal of Zhejiang University (Science Edition), 2023, 50 (06): 651-667.
[4] Wang Shibin, Gao Zidiao, Liu Dong An improved DCGAN image generation method based on limited data [J]. Journal of Henan Normal University (Natural Science Edition), 2023, 51 (06): 39-46.
[5] Xu Yongshi, Ben Kerong, Wang Tianyu, Liu Sijie Improvement of DCGAN Model and Research on SAR Image Generation [J]. Computer Science, 2020, 47 (12): 93-99.