基于知识蒸馏的轻量型浮游植物检测网络

doi:10.3969/j.issn.0255-8297.2020.03.003

摘要/Abstract

摘要： 当前基于卷积神经网络的目标检测框架已成为主流，使用深层的特征提取网络可以达到很好的目标检测效果，但带来的大量的参数和计算开销使这些算法难以应用到对存储空间和参数量有一定限制的嵌入式设备中.为此，该文提出将知识蒸馏方法用于目标检测网络的特征提取网络，以提升浅层特征提取网络的性能，在降低模型的计算量和规模的同时尽可能地保证模型的性能.实验结果表明，经过蒸馏的浅层网络作为特征提取网络的检测精度比没有经过教师指导的网络精度提高了11.7%.与此同时，该文构建的浮游植物目标检测数据集不仅可以评估一些最先进的目标检测算法的性能，也有利于未来浮游植物显微视觉技术的发展.

关键词: 知识蒸馏, 特征提取网络, 浮游植物, Faster R-CNN, 目标检测

Abstract: Object detection framework based on convolution neural network usually uses a very deep convolution neural network to extract object features before detection. However, its huge network structure leads to the reduction of detection speed, thus, the model can hardly achieve real-time object detection and be put into embedded devices. Address to the problem, this paper applies a knowledge distillation method to feature extraction network of object detection network to improve the performance of shallow feature extraction network. In this way, the model can ensure the same performance with a big reduction on computational load and model scale. Experimental results show that the detection accuracy of feature extraction networks employing distilled shallow network is 11.7% higher than that of networks without teacher’s guidance. Moreover, we build a phytoplankton dataset in this paper, which can not only be used for the evaluation of the performance of object detection algorithms, but also will be helpful to the development of phytoplankton microscopic vision technology.

Key words: knowledge distillation, feature extraction network, phytoplankton, Faster RCNN, object detection

中图分类号:

TP391

张彤彤, 董军宇, 赵浩然, 李琼, 孙鑫. 基于知识蒸馏的轻量型浮游植物检测网络[J]. 应用科学学报, 2020, 38(3): 367-376.

ZHANG Tongtong, DONG Junyu, ZHAO Haoran, LI Qiong, SUN Xin. Lightweight Phytoplankton Detection Network Based on Knowledge Distillation[J]. Journal of Applied Sciences, 2020, 38(3): 367-376.

参考文献

[1] Iandola F N, Han S, Moskewicz M W, et al. Squeezenet:AlexNet-level accuracy with 50x fewer parameters and <0.5 MB model size[C]//International Conference on Learning Representations, ICLR, Toulon, France, 2017:1-13.
[2] Hong S, Roh B, Kim K H, et al. Pvanet:lightweight deep neural networks for real-time object detection[C]//NIPS 2016 Workshop on Efficient Methods for Deep Neural Networks (EMDNN), Barcelona, Spain, 2016:1-7.
[3] Kim Y D, Park E, Yoo S, et al. Compression of deep convolutional neural networks for fast and low power mobile applications[J]. Computer Science, 2015, 71(2):576-584.
[4] Zhang X Y, Zou J H, He K M, et al. Accelerating very deep convolutional networks for classification and detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(10):1943-1955.
[5] Hinton G, Vinyals O, Dean J. Distilling the knowledge in a neural network[J]. Computer Science, 2015, 14(7):38-39.
[6] Bucilua C, Caruana R, Niculescu-Mizil A. Model compression[C]//Proceedings of the 12th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, New York, USA, 2006:535-541.
[7] Romero A, Ballas N, Kahou S E, et al. FitNets:hints for thin deep nets[J]. Computer Science, 2014(12):1-13.
[8] Ren S Q, He K M, Girshick, et al. Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2015, 39(6):1137-1149.
[9] He K M, Zhang X Y, Ren S Q, et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), IEEE Computer Society, Las Vegas, USA, 2016:770-778.
[10] 周飞燕,金林鹏,董军.卷积神经网络研究综述[J].计算机学报,2017(6):1229-1251. Zhou F Y, Jin L P, Dong J. Review of convolutional neural network[J]. Chinese Journal of Computers, 2017(6):1229-1251.(in Chinese)
[11] 张顺,龚怡宏,王进军.深度卷积神经网络的发展及其在计算机视觉领域的应用[J].计算机学报,2019, 42(3):3-32. Zhang S, Gong Y H, Wang J J. The development of deep convolution neural network and its applications on computer vision[J]. Chinese Journal of Computers, 2019, 42(3):3-32.(in Chinese)
[12] Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Columbus, USA, 2014:580-587.
[13] Girshick R. Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision, Santiago, Chile, 2015:1440-1448.
[14] Redmon J, Divvala S, Girshick R, et al. You only look once:unified, real-time object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, 2016:779-788.
[15] Liu W, Anguelov D, Erhan D, et al. SSD:single shot multibox detector[C]//European Conference on Computer Vision, Amsterdam, Holland, 2016:21-37.
[16] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[DB/OL]. 2014[2019-10-23]. https://arxiv.org/abs/1409.1556.
[17] 李涛,董前琨,张帅,等.基于线程池的GPU任务并行计算模式研究[J].计算机学报,2018, 41(10):2175-2192. Li T, Dong Q K, Zhang S, et al. GPU task parallel computing paradigm based on thread pool mode[J]. Chinese Journal of Computers, 2018, 41(10):2175-2192.(in Chinese)
[18] Urban G, Geras K J, Kahou S E, et al. Do deep convolutional nets really need to be deep and convolutional?[DB/OL]. 2016[2019-10-23].https://arxiv.org/pdf/1603.05691.
[19] Molchanov P, Tyree S, Karras T, et al. Pruning convolutional neural networks for resource efficient inference[C]//International Conference on Learning Representations, Toulon, France, 2017:1-17.
[20] Weinberger K, Dasgupta A, Langford J, et al. Feature hashing for large scale multitask learning[C]//Proceedings of the 26th Annual International Conference On Machine Learning, Montreal, Canada, 2009:1113-1120.
[21] Denil M, Shakibi B, Dinh L, et al. Predicting parameters in deep learning[C]//Advances in Neural Information Processing Systems, Lake Tahoe, Spain, 2013:2148-2156.
[22] Chen W L, Wilson J, Tyree S, et al. Compressing neural networks with the hashing trick[C]//International Conference on Machine Learning, Lille, France, 2015:2285-2294.
[23] Zhang X Y, Zou J H, Ming X, et al. Efficient and accurate approximations of nonlinear convolutional networks[C]//Proceedings of the IEEE Conference on Computer Vision and pattern Recognition, Boston, USA, 2015:1984-1992.
[24] Lecun Y, Denker J S, Solla S A. Optimal brain damage[C]//Advances in Neural Information Processing Systems, Denver, USA, 1990:598-605.
[25] Han S, Pool J, Tran J, et al. Learning both weights and connections for efficient neural network[C]//Advances in neural information processing systems, Montreal, Canada, 2015:1135-1143.
[26] Li H, Kadav A, Durdanovic I, et al. Pruning filters for efficient convNets[C]//International Conference on Learning Representations, Toulon, France, 2017:1-13.
[27] Lin X F, Zhao C, Pan W. Towards accurate binary convolutional neural network[C]//Advances in Neural Information Processing Systems, Long beach, USA, 2017:345-353.
[28] Lin Z, Courbariaux M, Memisevic R, et al. Neural networks with few multiplications[C]//International Conference on Learning Representations, San Juan, Puerto Rico, 2016:1-9.
[29] Gupta S, Agrawal A, Gopalakrishnan K, et al. Deep learning with limited numerical precision[C]//International Conference on Machine Learning, Lille, France, 2015:1737-1746.
[30] Zagoruyko S, Komodakis N. Paying more attention to attention:improving the performance of convolutional neural networks via attention transfer[C]//International Conference on Learning Representations, Toulon, France, 2017:1-13.
[31] Huang Z H, Wang N Y. Like what you like:knowledge distill via neuron selectivity transfer[C]//International Conference on Learning Representations, Toulon, France, 2017:1-9.
[32] Yim J, Joo D, Bae J, et al. A gift from knowledge distillation:fast optimization, network minimization and transfer learning[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Hawaii, USA, 2017:4133-4141.
[33] He K M, Zhang X Y, Ren S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9):1904-1916.
[34] Zeiler M D, Fergus R. Visualizing and understanding convolutional networks[C]//The 13th European Conference on Computer Vision, Zurich, Switzerland, 2014:818-833.