摘 要：深度学习算法在图像分类训练过程中存在着对训练样本需求量大以及难以获得较多样本等问题。文章对 VGG 网络模型的小样本学习问题进行研究，提出了基于数据增强和神经网络的小样本图像分类模型。改进模型通过主动数据增强学习结合DAGAN 进行数据增强，可在一定程度上缓解样本数量过少的问题，并在网络中引入注意力机制（VGG-SE），使得深度神经网络在小样本图像分类中达到较高的准确率，实验结果表明了所提模型的有效性。
关键词：小样本学习；图像分类；数据增强；神经网络
DOI:10.19850/j.cnki.2096-4706.2022.15.021
基金项目：芜湖市科技计划项目（2020yf48）；安徽理工大学环境友好材料与职业健康研究院研发专项基金资助项目（ALW2021YF04）
中图分类号：TP18 文献标识码：A 文章编号：2096-4706（2022）15-0077-04
Small Sample Image Classification Based on Data Augmentation and Neural Networks
YAN Jiajin1,2
(1.School of Computer Science and Engineering, Anhui University of Science and Technology, Huainan 232001, China; 2.Institute of Environment-friendly Materials and Occupational Health of Anhui University of Science and Technology (Wuhu), Wuhu 241003, China)
Abstract: In the process of image classification training, deep learning algorithm has many problems, such as large demand for training samples and difficult to obtain more samples. This paper studies the small sample learning problem of VGG network model, and proposes a small sample image classification model based on data enhancement and neural network. The improved model combines active data enhancement learning with DAGAN for data enhancement, which can alleviate the problem of too few samples to a certain extent, and introduce attention mechanism (VGG-SE) into the network, so that the deep neural network can achieve high accuracy in small sample image classification. The experimental results show the effectiveness of the proposed model.
Keywords: small sample learning; image classification; data enhancement; neural network
参考文献：
[1] 孙志军，薛磊，许阳明，等 . 深度学习研究综述 [J]. 计算机应用研究，2012，29（8）：2806-2810.
[2] 赵凯琳，靳小龙，王元卓 . 小样本学习研究综述 [J]. 软件学报，2021，32（2）：349-369.
[3] 刘颖，雷研博，范九伦，等 . 基于小样本学习的图像分类技术综述 [J]. 自动化学报，2021，47（2）：297-315.
[4] 祝钧桃，姚光乐，张葛祥，等 . 深度神经网络的小样本学习综述 [J]. 计算机工程与应用，2021，57（7）：22-33.
[5] HUISMAN M，VAN RIJN J N，Plaat A. A survey of deep meta-learning [J].Artificial Intelligence Review，2021，54（6）： 4483-4541.
[6] VINYALS O，BLUNDELL C，LILLICRAP T，et al. Matching networks for one shot learning [J].Advances in neural information processing systems，2016，29：3630-3638.
[7] HARIHARAN B，GIRSHICK R. Low-shot visual recognition by shrinking and hallucinating features [C]//2017 IEEE International Conference on Computer Vision (ICCV).Venice：IEEE，2017：3037-3049.
[8] RUSU A A，RAOD，SYGNOWSKI J，et al. Meta-Learning with Latent Embedding Optimization [J/OL].arXiv:1807.05960v3 [cs. LG].[2022-06-05].https://arxiv.org/abs/1807.05960v3.
[9] GOODFELLOW I J，POUGET-ABADIE J，MIRZA M，et al. Generative Adversarial Nets [J].Proceedings of the 27th International Conference on Neural Information Processing Systems，2014，2：2672-2680.
[10] ANTONIOU A，STORKEY A，E D WA R D S H . Data Augmentation Generative Adversarial Networks [J/OL]. arXiv:1711.04340v3 [stat.ML].[2022-06-08].https://arxiv.org/ abs/1711.04340.
[11] SIMONYAN K，ZISSERMAN A. Very deep convolutional networks for large-scale image recognition [J/OL].arXiv:1409.1556v6 [cs.CV].[2022-06-09].https://doi.org/10.48550/arXiv.1409.1556.
[12] 张宸嘉，朱磊，俞璐 . 卷积神经网络中的注意力机制综述 [J]. 计算机工程与应用，2021，57（20）：64-72.
[13] HU J，SHEN L，ALBANIE S. Squeeze-and-excitation networks [J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2018，42（8）：2011-2023.
作者简介：严家金（1998—），男，汉族，安徽安庆人，硕士研究生在读，研究方向：图像分类。