摘要
相比于GAN技术,手写数字生成技术没有那么亮眼,但它却是图像生成可以参考借鉴的最简单实例,从手写数字到手绘图像,甚至到手工雕像,人工智能已经渗透到了艺术创作领域并且通过训练创造出令人赞不绝口的艺术作品。主要研究了利用生成对抗网络进行手写数字生成的技术,基于对生成对抗网络原理的理解,在Tensor Flow深度学习框架与MNIST数据集的基础上,完成了网络的训练与测试,重点对比分析了MLP-GAN、DCGAN、CGAN以及C-DCGAN模型生成手写数字的情况,最后的训练结果表明,C-DCGAN生成的图片更清晰高效。
Compared to GAN,handwritten numeral generation technology is not as eye-catching,but it is the simplest example of image generation that can be referenced.From handwritten numerals to hand-painted images,even to handmade statues,artificial intelligence has infiltrated the field of artistic creation and created stunning works of art through training.This paper mainly studies the technology of generating handwritten numerals using generating confrontation networks.Based on an understanding of the principles of generating confrontation networks,training and testing of the network were completed on the basis of the TensorFlow deep learning framework and the MNIST dataset.A comparative analysis was focused on the generation of handwritten numerals using MLP-GAN,DCGAN,CGAN,and C-DCGAN models.The final training results show that the images generated by C-DCGAN are clearer and more efficient.
作者
丁泽云
Ding Zeyun(Yangzhou University,Yangzhou Jiangsu 225127)
出处
《现代工业经济和信息化》
2023年第4期263-265,共3页
Modern Industrial Economy and Informationization
