基于卷积神经网络的图像数据增强优化策略研究  

Research on Optimization Strategy of Image Data Enhancement Based on Convolutional Neural Network

作　　者：佟国香[1] 刘洪俊 田飞翔 TONG Guoxiang;LIU Hongjun;TIAN Feixiang(School of Optical Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《计算机与数字工程》2024年第7期2183-2188,共6页Computer & Digital Engineering

基　　金：国家重点研发计划项目(编号:2018YFB1700902)资助。

摘　　要：论文基于卷积神经网络模型,提出一种改进的自动化图像数据增强策略。针对原有策略搜索空间的离散化及模型训练过程中超参数优化不稳定的问题,通过降低搜索空间策略的复杂度、优化子网络模型的训练过程、选取更有效率的增强随机搜索算法实现超参数优化等方法对原有策略进行改进。并针对不同类型的数据集进行了验证,实验结果表明,论文提出的数据增强策略在CIFAR-10、CIFAR-100、ImageNet数据集上提升了图像分类的准确性,取得了先进的实验效果。Based on convolution neural network model,this paper proposes an improved automatic image data enhancement strategy.Focusing on the discretization of the original strategy search space and the instability of super parameter optimization in model training process,the original strategy is improved by reducing the complexity of search space strategy,optimizing the training process of sub-network model,and selecting more efficient enhanced random search algorithm to realize super parameter optimiza⁃tion.And the different types of datasets are used to verify.The results show that the proposed enhancement strategy improves the ac⁃curacy of image classification on CIFAR-10,CIFAR-100 and ImageNet datasets,it achieves advanced experimental results.

关 键 词：卷积神经网络 图像数据增强 超参数优化 增强随机搜索 图像分类 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]