基于深度迁移学习的中药饮片识别研究
|
摘要
由于传统的中药饮片识别与鉴定的准确性取决于操作者,存在个人主观性和不稳定性,故通过深度迁移学习对饮片进行辅助判断具有重大的应用价值。在Inception-V3模型的基础上依据finetune迁移方法对输出层进行变换,参考DDC(deep domain condusion)和DAN (deep adaptation networks)方法在模型中加入自适应度量来解决深度网络的自适应问题。通过测试集实验得到模型的识别率为88.3%,接近人工组实验的整体平均水平。
Because the accuracy of traditional Chinese medicine recognition and identification depends on the operator, and the operator has personal subjectivity and instability, it has significant application value that the judgment of decoction pieces assisted by deep migration learning. Based on the Inception-V3 model, this paper transforms the output layer according to the migration method Finetune, and solving the adaptive problem of deep network refers to DDC(deep domain condusion) and DAN method(deep adaptation networks) by adding adaptive metrics to the model. Through the test set experiment, the model recognition rate was 88.3%, which was close to the overall average of the artificial group.
Because the accuracy of traditional Chinese medicine recognition and identification depends on the operator, and the operator has personal subjectivity and instability, it has significant application value that the judgment of decoction pieces assisted by deep migration learning. Based on the Inception-V3 model, this paper transforms the output layer according to the migration method Finetune, and solving the adaptive problem of deep network refers to DDC(deep domain condusion) and DAN method(deep adaptation networks) by adding adaptive metrics to the model. Through the test set experiment, the model recognition rate was 88.3%, which was close to the overall average of the artificial group.
引文
[1] SHEN D G, WU G R, SUK H. Deep learning in medical image analysis[J]. Annual review of biomedical engineering,2017(19):221-248.
[2] BRUNO O M, PLOTZE R O, FALVO M, et al. Fractal dimension applied to plant identification[J]. Information sciences, 2008, 12(178):2722-2733.
[3]侯铜,姚立红,阚江明.基于叶片外形特征的植物识别研究[J].湖南农业科学,2009(4):123-125.
[4]陶欧,林兆洲,张宪宝,等.基于饮片切面图像纹理特征参数的中药辨识模型研究[J].世界科学技术:中医药现代化,2014(12):2558-2562.
[5] YOSINSKI J, CLUNE J, BENGIO Y, et al. How transferable are features in deep neural networks?[C]//NIPS, Advances in Neural Information Processing Systems, December 8-13, 2014, Montréal. New York:NIPS, 2014:3320-3328.
[6] DONAHUE J, JIA Y Q, VINYALS O, et al. A deep convolutional activation feature for generic visual recognition[C]//ASCE, International Conference on Machine Learning and Cybernetice, July 13-16, 2014, Lanzhou.NewYork:ASCE, 2014:647-655.
[7] PAN S J, YANG Q. A survey on transfer learning[J].IEEE transactions on knowledge and data engineering,2010, 22(10):1345-1359.
[8] TAN B, SONG Y Q, ZHONG E, et al. Transitive transfer learning[C]//ACM, Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, August 10-13, 2015, Sydney. NewYork:ACM, 2015:1155-1164.
[9] TZENG E, HOFFMAN J, ZHANG N, et al. Deep domain confusion:Maximizing for domain invariance[C]//IEEE,Computer Vision and Pattern Recognition. June 24-27,2014, Columbus. Washington D. C.:IEEE, 2014:1412-1434.
[10] LONG M S, CAO Y, WANG J M, et al. Learning transferable features with deep adaptation networks[C]//ACM, Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. August 10-13, 2015, Sydney. NewYork:ACM, 2015:97-105.
[2] BRUNO O M, PLOTZE R O, FALVO M, et al. Fractal dimension applied to plant identification[J]. Information sciences, 2008, 12(178):2722-2733.
[3]侯铜,姚立红,阚江明.基于叶片外形特征的植物识别研究[J].湖南农业科学,2009(4):123-125.
[4]陶欧,林兆洲,张宪宝,等.基于饮片切面图像纹理特征参数的中药辨识模型研究[J].世界科学技术:中医药现代化,2014(12):2558-2562.
[5] YOSINSKI J, CLUNE J, BENGIO Y, et al. How transferable are features in deep neural networks?[C]//NIPS, Advances in Neural Information Processing Systems, December 8-13, 2014, Montréal. New York:NIPS, 2014:3320-3328.
[6] DONAHUE J, JIA Y Q, VINYALS O, et al. A deep convolutional activation feature for generic visual recognition[C]//ASCE, International Conference on Machine Learning and Cybernetice, July 13-16, 2014, Lanzhou.NewYork:ASCE, 2014:647-655.
[7] PAN S J, YANG Q. A survey on transfer learning[J].IEEE transactions on knowledge and data engineering,2010, 22(10):1345-1359.
[8] TAN B, SONG Y Q, ZHONG E, et al. Transitive transfer learning[C]//ACM, Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, August 10-13, 2015, Sydney. NewYork:ACM, 2015:1155-1164.
[9] TZENG E, HOFFMAN J, ZHANG N, et al. Deep domain confusion:Maximizing for domain invariance[C]//IEEE,Computer Vision and Pattern Recognition. June 24-27,2014, Columbus. Washington D. C.:IEEE, 2014:1412-1434.
[10] LONG M S, CAO Y, WANG J M, et al. Learning transferable features with deep adaptation networks[C]//ACM, Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. August 10-13, 2015, Sydney. NewYork:ACM, 2015:97-105.