一种基于生成对抗网络的行为数据集扩展方法
|
摘要
深度学习作为人工神经网络的分支,在图像识别领域有广泛的应用,但其数据集的不足导致模型学习不够完善。通过对深度学习的数据规模要求进行分析,针对人体行为识别中的应用,发现人体数据集的采集工作是一个极具耗时耗力的工程,很难满足目前深度学习网络的需求。为了解决这一难题,提出了一种依靠原有的小规模数据集产生大量可靠数据集的半监督深度学习模型。通过将循环神经网络和生成式对抗网络相结合的方法使循环神经网络学习到数据的序列关系和特征,使生成式对抗网络产生合理数据进而扩展人体行为数据集。依靠该网络结构,可以很好地分析出采集数据的特征,并且依据这些特征可以生成大量的合理的数据,后经过数据处理等工作,形成可用于模型训练的可靠数据集,缓解了深度学习工作中数据集紧缺的问题。
As a branch of artificial neural network,deep learning has a wide range of applications in the field of image recognition. The lack of data sets leads to incomplete model learning. Through the analysis of the data size requirements of deep learning,it is found that the collection of human data sets is a very time-consuming and labor-intensive project for the application of human behavior recognition. It is difficult to meet the needs of the current deep learning network. To solve this problem,we propose a semi-supervised deep learning model that relies on the original small-scale data set to generate a large number of reliable data sets. By combining the cyclic neural network and the generative confrontation network,the cyclic neural network learns the sequence relationship and characteristics of the data,so that the generation-oriented network generates reasonable data and then expands the human behavior data set. Relying on this network structure,the characteristics of the collected data can be well analyzed,and a large amount of reasonable data can be generated according to these features,and then processed through data processing to form a reliable data set that can be used for model training,thereby alleviating the shortage of data sets in deep learning work.
As a branch of artificial neural network,deep learning has a wide range of applications in the field of image recognition. The lack of data sets leads to incomplete model learning. Through the analysis of the data size requirements of deep learning,it is found that the collection of human data sets is a very time-consuming and labor-intensive project for the application of human behavior recognition. It is difficult to meet the needs of the current deep learning network. To solve this problem,we propose a semi-supervised deep learning model that relies on the original small-scale data set to generate a large number of reliable data sets. By combining the cyclic neural network and the generative confrontation network,the cyclic neural network learns the sequence relationship and characteristics of the data,so that the generation-oriented network generates reasonable data and then expands the human behavior data set. Relying on this network structure,the characteristics of the collected data can be well analyzed,and a large amount of reasonable data can be generated according to these features,and then processed through data processing to form a reliable data set that can be used for model training,thereby alleviating the shortage of data sets in deep learning work.
引文
[1] NGUYEN L T,NGUYEN N T,VO B,et al.Efficient method for updating class association rules in dynamic datasets with record deletion[J].Applied Intelligence,2018,48(6):1491-1505.
[2] WEI Wei.Information retrieval in biomedical research:from articles to datasets[D].San Diego:University of California,2017.
[3] HINTON G E,OSINDERO S,TEH Y W.A fast learning algorithm for deep belief nets[J].Neural Computation,2006,18(7):1527-1554.
[4] 郑胤,陈权崎,章毓晋.深度学习及其在目标和行为识别中的新进展[J].中国图象图形学报,2014,19(2):175-184.
[5] 裴晓敏,范慧杰,唐延东.时空特征融合深度学习网络人体行为识别方法[J].红外与激光工程,2018,47(2):0203007-1-0203007-6.
[6] 单言虎,张彰,黄凯奇.人的视觉行为识别研究回顾、现状及展望[J].计算机研究与发展,2016,53(1):93-112.
[7] SHAIKHINA T,KHOVANOVA N A.Handling limited datasets with neural networks in medical applications:a small-data approach[J].Artificial Intelligence in Medicine,2017,75:51-63.
[8] SCHULDT C,LAPTEV I,CAPUTO B.Recognizing human actions:a local SVM approach[C]//International conference on pattern recognition.Cambridge:IEEE,2004:32-36.
[9] LAPTEV I.Local spatio-temporal image features for motion interpretation[D].Stockholm:Royal Institute of Technology,2004.
[10] LAPTEV I,LINDEBERG T.Velocity adaptation of space-time interest points[C]//International conference on pattern recognition.Cambridge:IEEE,2004:52-56.
[11] WEINLAND D,BOYER E,RONFARD R.Action recognition from arbitrary views using 3D exemplars[C]//International conference on computer vision.Rio de Janeiro:IEEE,2007:1-7.
[12] ZHANG Zhang,TAO Dacheng.Slow feature analysis for human action recognition[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2012,34(3):436-450.
[13] MARSZALEK M,LAPTEV I,SCHMID C.Actions in context[C]//IEEE conference on computer vision and pattern recognition.Miami,FL,USA:IEEE,2009:2929-2936.
[14] LIU Jingen,LUO Jiebo,SHAH M.Recognizing realistic actions from videos[C]//IEEE conference on computer vision and pattern recognition.Miami,FL,USA:IEEE,2009:1996-2003.
[15] SOOMRO K,ZAMIR A R,SHAH M.UCF101:a dataset of 101 human actions classes from videos in the wild[J].Computer Science,2012,4(2):1212-1219.
[16] 张营营.生成对抗网络模型综述[J].电子设计工程,2018,26(5):34-37.
[17] CHO K,VAN MERRIENBOER B,GULCEHRE C,et al.Learning phrase representations using RNN encoder-decoder for statistical machine translation[C]//Empirical methods on natural language processing.[s.l.]:[s.n.],2014.
[18] 赵树阳,李建武.基于生成对抗网络的低秩图像生成方法[J].自动化学报,2018,44(5):829-839.
[19] 曹志义,牛少彰,张继威.基于半监督学习生成对抗网络的人脸还原算法研究[J].电子与信息学报,2018,40(2):323-330.
[20] SUNDERMEYER M,SCHLüTER R,NEY H.LSTM neural networks for language modeling[EB/OL].[2014-02-10].http://www-i6.informatik.rwth-aachen.de/publications/download/820/Sundermeyer-2012.pdf.
[21] GIRSHICK R.Fast R-CNN[J].Computer Science,2015,7(3):1356-1361.
[22] RAZAVIAN A S,AZIZPOUR H,SULLIVAN J,et al.CNN features off-the-shelf:an astounding baseline for recognition[C]//IEEE conference on computer vision and pattern recognition workshops.[s.l.]:IEEE,2014:806-813.
[2] WEI Wei.Information retrieval in biomedical research:from articles to datasets[D].San Diego:University of California,2017.
[3] HINTON G E,OSINDERO S,TEH Y W.A fast learning algorithm for deep belief nets[J].Neural Computation,2006,18(7):1527-1554.
[4] 郑胤,陈权崎,章毓晋.深度学习及其在目标和行为识别中的新进展[J].中国图象图形学报,2014,19(2):175-184.
[5] 裴晓敏,范慧杰,唐延东.时空特征融合深度学习网络人体行为识别方法[J].红外与激光工程,2018,47(2):0203007-1-0203007-6.
[6] 单言虎,张彰,黄凯奇.人的视觉行为识别研究回顾、现状及展望[J].计算机研究与发展,2016,53(1):93-112.
[7] SHAIKHINA T,KHOVANOVA N A.Handling limited datasets with neural networks in medical applications:a small-data approach[J].Artificial Intelligence in Medicine,2017,75:51-63.
[8] SCHULDT C,LAPTEV I,CAPUTO B.Recognizing human actions:a local SVM approach[C]//International conference on pattern recognition.Cambridge:IEEE,2004:32-36.
[9] LAPTEV I.Local spatio-temporal image features for motion interpretation[D].Stockholm:Royal Institute of Technology,2004.
[10] LAPTEV I,LINDEBERG T.Velocity adaptation of space-time interest points[C]//International conference on pattern recognition.Cambridge:IEEE,2004:52-56.
[11] WEINLAND D,BOYER E,RONFARD R.Action recognition from arbitrary views using 3D exemplars[C]//International conference on computer vision.Rio de Janeiro:IEEE,2007:1-7.
[12] ZHANG Zhang,TAO Dacheng.Slow feature analysis for human action recognition[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2012,34(3):436-450.
[13] MARSZALEK M,LAPTEV I,SCHMID C.Actions in context[C]//IEEE conference on computer vision and pattern recognition.Miami,FL,USA:IEEE,2009:2929-2936.
[14] LIU Jingen,LUO Jiebo,SHAH M.Recognizing realistic actions from videos[C]//IEEE conference on computer vision and pattern recognition.Miami,FL,USA:IEEE,2009:1996-2003.
[15] SOOMRO K,ZAMIR A R,SHAH M.UCF101:a dataset of 101 human actions classes from videos in the wild[J].Computer Science,2012,4(2):1212-1219.
[16] 张营营.生成对抗网络模型综述[J].电子设计工程,2018,26(5):34-37.
[17] CHO K,VAN MERRIENBOER B,GULCEHRE C,et al.Learning phrase representations using RNN encoder-decoder for statistical machine translation[C]//Empirical methods on natural language processing.[s.l.]:[s.n.],2014.
[18] 赵树阳,李建武.基于生成对抗网络的低秩图像生成方法[J].自动化学报,2018,44(5):829-839.
[19] 曹志义,牛少彰,张继威.基于半监督学习生成对抗网络的人脸还原算法研究[J].电子与信息学报,2018,40(2):323-330.
[20] SUNDERMEYER M,SCHLüTER R,NEY H.LSTM neural networks for language modeling[EB/OL].[2014-02-10].http://www-i6.informatik.rwth-aachen.de/publications/download/820/Sundermeyer-2012.pdf.
[21] GIRSHICK R.Fast R-CNN[J].Computer Science,2015,7(3):1356-1361.
[22] RAZAVIAN A S,AZIZPOUR H,SULLIVAN J,et al.CNN features off-the-shelf:an astounding baseline for recognition[C]//IEEE conference on computer vision and pattern recognition workshops.[s.l.]:IEEE,2014:806-813.