| 502 | 48 | 28 |
| 下载次数 | 被引频次 | 阅读次数 |
当前,在图像目标检测识别方面,深度学习技术已经成为研究的热点。然而深度学习在进行网络训练时需要使用大量的样本,当样本数目较少时,得到的训练模型其检测效果往往不佳。介绍了色彩变换、水平翻转、旋转、亮度变换、缩放、裁剪、添加噪声等不同数据增强方法,并结合VOC2007数据集,采用数据增强技术实现样本扩充。实验结果表明对样本进行数据增强处理,均可以在一定程度上提高图像检测精度。特别是色彩变换、水平翻转、旋转、亮度变换和缩放这五种方法扩展训练集,对最终检测效果提升较为明显。
Abstract：At present, deep learning technology has become a hot research direction in image target detection and recognition.However, a large number of samples are needed for network training in deep learning.When the number of samples is small, the detection effect of the training model obtained is often poor.In this paper, different data enhancement methods such as color transformation, horizontal flip, rotation, brightness transformation, scaling, clipping and adding noise are introduced.Combined with VOC2007 data set, data enhancement technology is adopted to realize sample expansion.The experimental results show that the image detection accuracy can be improved to some extent through data enhancement processing.In particular, the five methods of color transformation, horizontal reversal, rotation, brightness transformation and scaling can expand the training set, which improve the detection effect significantly.
[1] HE K M,ZHANG X Y,REN S Q,et al.Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition.Las Vegas,NV,USA:IEEE,2016:770-778.
[2] HINTON G E.,OSINDERO S,TEH Y W.A fast learning algorithm for deep belief nets[J].Neural Computation,2006,18(7):1527-1554.
[3] SCHMIDHUBER J.Deep learning in neural networks:an overview[J].Neural Networks,2015,61(4):85-117.
[4] LEE H,PHAM P,LARGMAN Y,et al.Unsupervised feature learning for audio classification using convolutional deep belief networks[C]//Advances in Neural Information Processing Systems,2009:1096-1104.
[5] DESELAERS T,HASAN S,BENDER O,et al.A deep learning approach to machine transliteration[C]//Proceedings of the Fourth Workshop on Statistical Machine Translation,2009:233-241.
[6] GLOROT X,BORDES A,BENGIO Y.Domain adaptation for large-scale sentiment classification:A deep learning approach[C]//ICML,2011:1024-1028.
[7] RANZATO M A,SUSSKIND J,MNIH V,et al.On deep generative models with applications to recognition.[C]//CVPR 2011.IEEE,2011:2857-2864.
[8] LEE H,EKANADHAM C,NG A Y.Sparse deep belief net model for visual area V2[C]//Advances in Neural Information Processing Systems,2007:873-880.
[9] TAYLOR G W,HINTON G E,ROWEIS S T.Modeling human motion using binary latent variables[C]//Advances in Neural Information Processing Systems,2007:1345-1352.
[10] LIU W,ANGUELOWD,ERHAN D,et al.SSD:Single shot multibox detector[C]//European Conference on Computer Vision.Springer,Cham,2016:21-37.
[11] GIRSHICK R,DONAHUE J,DARRELL T,et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//IEEE Conference on Computer Vision and Pattern Recognition,2014:580-587.
[12] REDMON J,DIVVALA S,GIRSHICK R,et al.You only look once:unified,real-time object detection[C]//IEEE Conference on Computer Vision and Pattern Recognition,2016:779-788.
[13] 黄长专，王彪，杨忠.图像分割方法研究[J].计算机技术与发展，2009,19(6):76-79.
[14] KRIZHEVSKY A,SUTSKEVERI,HINTON G.Imagenet classification with deep convolutional neural networks[J].Communications of The ACM,2017,60(6):84-90.
[15] SMITH A R.Color gamut transform pairs[J].ACM Siggraph Computer Graphics,1978,12(3):12-19.
[16] 陈文兵，管正雄，陈允杰.基于条件生成式对抗网络的数据增强方法[J].计算机应用，2018,38(11):3305-3311.
[17] GOODFELLOW I,POUGET-ABADIE J,MIRZA M,et al.Generative adversarial nets[C]//Advances in Neural Information Processing Systems,2014:2672-2680.
[18] TAKAHASHI R,MATSUBARA T,UEHARA K R.Random image cropping and patching data augmentation for deep cnns[C]//Asian Conference on Machine Learning,2018:786-798.
基本信息:
DOI：10.16426/j.cnki.jcdzdk.2021.01.014
中图分类号:TP391.41;TP18
引用信息:
[1]李永盛,何佳洲,刘义海,等.基于图像检测识别的数据增强技术[J].舰船电子对抗,2021,44(01):66-70.DOI:10.16426/j.cnki.jcdzdk.2021.01.014.