SealGAN: 基于生成式对抗网络的印章消除研究

发票是财务系统的重要组成部分. 随着计算机视觉和人工智能技术的发展, 出现了各种发票自动识别系统, 但是发票上的印章严重影响了识别准确率. 本文提出了一种用于自动消除发票印章的SealGAN网络. SealGAN网络是基于生成式对抗网络CycleGAN的改进, 采用两个独立的分类器来取代原本的判别网络, 从而降低单个分类器的分类要求, 提高分类器的学习性能, 并且结合ResNet和Unet两种结构构建下采样?精炼?上采样的生成网络, 生成更加清晰的发票图像. 同时提出了基于风格评价和内容评价的综合评价指标对SealGAN网络进行性能评价. 实验结果表明, 与CycleGAN-ResNet和CycleGAN-Unet网络相比较, Seal GAN网络不仅能实现自动消除印章, 而且还能更加清晰地保留印章下的发票内容, 网络性能评价指标较高.

SealGAN: Research on the Seal Elimination Based on Generative Adversarial Network

Invoice is an important part of the financial system. With the development of computer vision and artificial intelligence technologies, various automatic invoice identification systems have been developed. However, the seals on the invoice often affect the identification success rate. In this paper, the SealGAN network structure is proposed, which can automatically eliminate the seal of invoice. SealGAN network is an improvement based on generative adversarial network CycleGAN. The two discriminant networks are replaced with two independent classifiers in the SealGAN, which reduces the classification requirement of each classifier, and the learning performance of classifiers can be improved. In addition, it combines the ResNet and UNet structures to construct a downsampling-refining-upsampling generation network, which can generate clearer images of invoice. And the network comprehensive evaluation index, including style evaluation and content evaluation, is proposed to evaluate the performance of network. Experimental results demonstrate that the SealGAN network can not only eliminate the seal automatically, but also retain the invoice content under the seal clearly, and the network performance evaluation index is higher than that of the CycleGAN-ResNet and CycleGAN-Unet.