|
|
改进生成对抗网络的小样本图像数据增强算法
|
Abstract:
针对小样本场景下深度学习图像数据集匮乏,且现有生成式对抗网络(Generative Adversarial Networks, GAN)存在生成图像清晰度低、训练稳定性差的问题,提出一种融合相对判别器与零中心梯度惩罚机制(R1,R2正则化)的相对论配对生成对抗网络(Relativistic pairing GAN, RpGAN)图像数据增强算法。算法对GAN网络进行改进,通过相对判别器捕捉真实样本与生成样本的相对差异,缓解模式坍塌;引入R1,R2零中心梯度惩罚机制,保证模型局部收敛性,提升训练稳定性;在生成器中嵌入残差块,强化小样本特征提取能力,提升生成图像清晰度。实验结果表明,该方法用于MNIST及耐张线夹X光小样本数据集时,图像增强效果FID值对比未经改进的网络分别降低24.3%与29.7%。
In response to the scarcity of deep learning image datasets in small-sample scenarios, and the issues of low image clarity and poor training stability in existing Generative Adversarial Networks (GAN), we propose a Relativistic pairing GAN (RpGAN) image data augmentation algorithm that integrates a relative discriminator and a zero-centered gradient penalty mechanism (R1, R2 regularization). This algorithm improves the GAN network by capturing the relative differences between real and generated samples through a relative discriminator, mitigating mode collapse. It introduces the R1 and R2 zero-centered gradient penalty mechanisms to ensure local convergence of the model and enhance training stability. Additionally, residual blocks are embedded in the generator to strengthen the ability to extract small-sample features and improve the clarity of generated images. Experimental results show that when applied to the MNIST and tension clamp X-ray small-sample datasets, the image enhancement effect shows that the FID values are reduced by 24.3% and 29.7% compared to the unimproved network, respectively.
| [1] | 张晓峰, 吴刚. 基于生成对抗网络的数据增强方法[J]. 计算机系统应用, 2019, 28(10): 201-206. |
| [2] | 徐海青, 余江斌, 梁翀, 张晓航. 基于GAN的改进RPN输电线路细小金具缺陷检测方法[J]. 电子器件, 2021, 44(6): 1409-1416. |
| [3] | Goodfellow, I.J., Pouget Abadie, J., Mirza, M., et al. (2014) Generative Adversarial Networks. arXiv: 1406.2661 |
| [4] | 曹仰杰, 贾丽丽, 陈永霞, 等. 生成式对抗网络及其计算机视觉应用研究综述[J]. 中国图象图形学报, 2018, 23(10): 1433-1449. |
| [5] | Sun, R.Y., Fang, T.T. and Schwing, A. (2020) Towards a Better Global Loss Landscape of GANs. arXiv: 2011.04926 |
| [6] | Jolicoeur Martineau, A. (2019) The Relativistic Discriminator: A Key Element Missing from Standard GAN. arXiv: 1807.00734 |
| [7] | Mescheder, L., Geiger, A. and Nowozin, S. (2018) Which Training Methods for GANs Do Actually Converge? arXiv: 1801.04406 |
| [8] | Gulrajani, I., Ahmed, F., Arjovsky, M., et al. (2017) Improved Training of Wasserstein GANs. arXiv: 1704.00028 |
| [9] | Roth, K., Lucchi, A., Nowozin, S., et al. (2017) Stabilizing Training of Generative Adversarial Networks through Regularization. arXiv: 1705.09367 |
| [10] | 郭伟, 庞晨. 改进生成式对抗网络的图像数据集增强算法[J]. 电讯技术, 2022, 62(3): 281-287. |
| [11] | Chen, L. and Wang, H. (2022) Improved Relativistic Pairing GAN with Residual Blocks for Few Shot Image Generation. IEEE International Conference on Signal Processing, Beijing, 21-24 October 2022, 1345-1349. |
| [12] | Radford, A., Metz, L. and Chintala, S. (2016) Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks. arXiv: 1511.06434. https://arxiv.org/abs/1511.06434 |
| [13] | 杨东升, 吉明佳, 周博文, 等. 基于双生成器生成对抗网络的电力系统暂态稳定评估方法[J]. 电网技术, 2021, 45(08): 2934-2944. |
| [14] | 段亚茹. 基于生成对抗网络的文本生成图像算法[D]: [硕士学位论文]. 西安: 长安大学, 2023. |
| [15] | Karras, T., Laine, S., Aittala, M., et al. (2020) Analyzing and Improving the Image Quality of StyleGAN. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, 13-19 June 2020, 8110-8119. https://doi.org/10.1109/CVPR42600.2020.00813 |
| [16] | 吴少乾, 李西明. 生成对抗网络的研究进展综述[J]. 计算机科学与探索, 2020, 14(3): 377-388. |
| [17] | He, K., Zhang, X., Ren, S. and Sun, J. (2016) Deep Residual Learning for Image Recognition. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, 27-30 June 2016, 770-778. https://doi.org/10.1109/cvpr.2016.90 |
| [18] | 王星, 杜伟, 陈吉, 等. 基于深度残差生成式对抗网络的样本生成方法[J]. 控制与决策, 2020, 35(8): 1887-1894. |
| [19] | 徐贵力, 刘小霞, 田裕鹏, 等. 一种图像清晰度评价方法[J]. 红外与激光工程, 2009, 38(1): 180-184. |
| [20] | Chen, J., Zhang, Y., Bai, S., Zhu, J., Chirarattananon, P., Ni, K., et al. (2021) Dual-Foci Fast-Scanning Photoacoustic Microscopy with 3.2-MHz A-Line Rate. Photoacoustics, 23, Article ID: 100292. https://doi.org/10.1016/j.pacs.2021.100292 |
| [21] | Heusel, M., Ramsauer, H., Unterthiner, T., et al. (2017) GANs Trained by a Two Time Scale Update Rule Converge to a Local Nash Equilibrium. Proceedings of the 31st International Conference on Neural Information Processing Systems, Long Beach, 4-9 December 2017, 6629-6640. |
| [22] | Zhang, H., Goodfellow, I., Metaxas, D., et al. (2019) Self Attention Generative Adversarial Networks. Proceedings of the 36th International Conference on Machine Learning, Long Beach, 9-15 June 2019, 7354 7363. |
| [23] | Lucic, M., Kurach, K., Michalski, M., et al. (2018) Are GANs Created Equal? A Large Scale Study. arXiv: 1711.10337 |
| [24] | 董永生, 范世朝, 张宇, 等. 生成对抗网络的发展与挑战[J]. 信号处理, 2023, 39(1): 154-175. |
| [25] | Xia, X.L., Xu, C. and Nan, B. (2017) Inception-v3 for Flower Classification. 2017 2nd International Conference on Image, Vision and Computing (ICIVC), Chengdu, 2-4 June 2017, 783-787. https://doi.org/10.1109/icivc.2017.7984661 |