Abstract
Convolutional neural network (CNN) has emerged as one of the primary tools for image classification. In particular, deep CNNs are now considered the main tool for this purpose. However, these networks are often large and require computing and storage power that may not be available in very small sensor devices such as IoT (Internet of Things) devices; their training is also time and computing power consuming. As a result, in some applications, reducing the size of input data (images) and the processing network becomes necessary. Such reduction usually comes at the cost of reduced classification performance. In this paper, we consider networks with under 200k learnable parameters, as opposed to millions in deeper networks. We examine how domain transforms can be used for efficient size reduction and improvement of classification accuracy for small networks. We emphasize that finding optimal hyperparameters or network configurations is not our objective in this paper. It is shown that by using transforms such as discrete wavelet transforms (DWT) or discrete cosine transform (DCT), it is possible to efficiently improve the performance of size-reduced networks and inputs. We demonstrate that in most cases, the improvement can be traced to higher entropy of resized input using transforms. While transforms such as DCT allow variable input and network sizes to be utilized, DWT proves to be very effective when significant size reduction is needed (improving the result by up to 5%). It is also shown that input size reduction of up to 75% is possible, without loss of classification accuracy in some cases. We use two datasets of small images, including Fashion MNIST and CIFAR-10, to evaluate the performance of size reduction methods.
Similar content being viewed by others
Code Availability
MATLAB neural network toolbox is used in this research, which is commercially available.
References
R. Asadi, S.A. Kareem, Review of Feed Forward Neural Network classification preprocessing techniques. In Proceedings of the 3rd Int. Conf. on Math. Sciences, AIP Conf. Proc., vol. 1602 (2014), p. 567
Al. Aussem, F. Murtagh, combining neural network forecasts on wavelet-transformed time series. Connect. Sci. 9, 113–122 (1997)
A. Ciurana, A. Mosella-Montoro, J. Ruiz-Hidalgo, Hybrid cosine based convolutional neural networks (2019). https://arxiv.org/pdf/1904.01987.pdf
A. Coates, H. Lee, A. Ng, An analysis of single-layer networks in unsupervised feature learning. In NIPS Workshop on Deep Learning (2010)
A. Cohen, I. Daubechies, J.-C. Feauveau, Biorthogonal bases of compactly supported wavelets. Commun. Pure Appl. Math. 45(5), 485–560 (1992). https://doi.org/10.1002/cpa.3160450502
I. Daubechies, Ten lectures on wavelets. SIAM (1992). https://doi.org/10.1137/1.9781611970104
D. Fu, G. Guimaraes, Using compression to speed up image classification in artificial neural networks “the discrete cosine transforms” (2016), pp. 1–10
A. Ghosh, R. Chellappa, Deep feature extraction in the DCT domain. In 23rd International Conference on Pattern Recognition (ICPR), pp. 3525–3530 (2016)
B. Graham, :Fractional max-pooling. In arxiv:cs/arXiv:1412.6071, 2015.
M. Hashemi, Enlarging smaller images before inputting into the convolutional neural network: zero-padding vs interpolation. J. Big Data 6, 98 (2019). https://doi.org/10.1186/s40537-019-0263-7
K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition (2015). arXiv:1512.03385
L. Hertel, E. Barth, T. Käster, T. Martinetz, Deep convolutional neural networks as generic feature extractors. In IJCNN (2015)
https://www.mathworks.com/solutions/deep-learning/examples/training-a-model-from-scratch.html
D. Huang, X.-R. Bai, A wavelet neural network optimal control model for traffic-flow prediction in intelligent transport systems. Adv. intelligent computing theories and apps. with aspects of A.I.. ICIC 2007. In LNCS, vol. 4682 (2007), pp. 1233–1244
X. Jiang, H. Adeli, Dynamic wavelet neural network model for traffic flow forecasting. J. Transport. Eng. 131(10), 771–779 (2005)
M. Kayed, A. Anter, H. Mohamed,: Classification of garments from fashion MNIST dataset using CNN LeNet-5 architecture. In Int. Conf. on Innovative Trends in Communication and Computer Engineering (ITCE) (2020), pp. 238–243. https://doi.org/10.1109/ITCE48509.2020.9047776
M.K. Khandani, W.B. Mikhael, Using mixed DCT and Haar transforms for efficient compression of car trajectory data. In 2018 IEEE 61st International Midwest Symposium on Circuits and Systems (MWSCAS) (2018), pp. 692–695. https://doi.org/10.1109/MWSCAS.2018.8623987
M.K. Khandani, W.B. Mikhael, Efficient time series forecasting using time delay neural networks with domain pre-transforms. In 2019 IEEE 62nd International Midwest Symposium on Circuits and Systems (MWSCAS) (2019), pp. 682–685. https://doi.org/10.1109/MWSCAS.2019.8884826
M.K. Khandani, W.B. Mikhael, A study on network size reduction using sparse input representation in time delay neural networks. In 2020 IEEE 63rd International Midwest Symposium on Circuits and Systems (MWSCAS) (2020), pp. 864–867. https://doi.org/10.1109/MWSCAS48704.2020.9184438
M.K. Khandani, W.B. Mikhael, Effect of sparse representation of time series data on learning rate of time delay neural networks. Circuits Syst. Signal Process. (2021). https://doi.org/10.1007/s00034-020-01610-8
A. Krizhevsky. Learning multiple layers of features from tiny images. Technical report (2009)
A. Krizhevsky, I. Sutskever, G.E. Hinton, ImageNet classification with deep convolutional neural networks. Commun. ACM 60(6), 84–90 (2017). https://doi.org/10.1145/3065386
J. Langelaar, MNIST neural network training and testing (https://www.mathworks.com/matlabcentral/fileexchange/73010-mnist-neural-network-training-and-testing), MATLAB Central File Exchange. Retrieved December 11, 2020
B. Lee, Y.S. Tarng, Application of the discrete wavelet transform to the monitoring of tool failure in end milling using the spindle motor current. Int. J. Adv. Manuf. Technol. (1999). https://doi.org/10.1007/s001700050062
P. Liu, H. Zhang, W. Lian, W. Zuo, Multi-level wavelet convolutional neural networks: https://arxiv.org/pdf/1907.03128.pdf
Z. Pan, H. Bolouri, High speed face recognition based on discrete cosine transforms and neural networks. http://citeseer.ist.psu.edu/270448.html. Submitted to IEEE Trans. On PAMI 1999
B. Rajesh, M. Javed, Ratnesh, S. Srivastava, DCT-CompCNN: a novel image classification network using JPEG compressed DCT coefficients
S. Said, O. Jemai, S. Hassairi, R. Ejbali, M. Zaied, C.B. Amar, Deep wavelet network for image classification. In 2016 IEEE International Conference on Systems, Man, and Cybernetics (20160, pp. 922–927
S. Shubathra, P.C.D. Kalaivaani, S. Santhoshkumar, Clothing image recognition based on multiple features using deep neural networks. In International Conference on Electronics and Sustainable Communication Systems (ICESC) (IEEE , 2020)
K. Simonyan, A. Zisserman, Very deep convolutional networks for large-scale image recognition, 2015, arXiv:1409.1556v6
J.T. Springenberg, A. Dosovitskiy, T. Brox, M. Riedmiller, Striving for simplicity: the all convolutional net, workshop contribution at ICLR 2015, arXiv:1412.6806v3
C. Thum, Measurement of the entropy of an image with application to image focusing (2010), pp. 203–211. https://doi.org/10.1080/713821475
A. Torralba, R. Fergus, W. Freeman, 80 million tiny images: A large data set for nonparametric object and scene recognition. IEEE PAMI 30(11), 1958–1970 (2008)
Y. Wang, C. Xu, C. Xu, D. Tao, Packing convolutional neural networks in the frequency domain. IEEE Trans. Pattern Anal. Mach. Intel. 41(10), 2495–2510 (2019)
T. Williams, R. Li, Advanced image classification using wavelets and convolutional neural networks. In 2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA) (2016), pp. 233–239
H. Xiao, K. Rasul, R. Vollgraf, Fashion-MNIST: a novel image dataset for benchmarking machine learning algorithms (2017). https://arxiv.org/pdf/1708.07747.pdf
B. Zhou, A. Khosla, A. Lapedriza, A. Oliva, A. Torralba: Learning deep features for discriminative localization. In CVPR (2016)
Funding
Not Applicable.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Kalantari Khandani, M., Mikhael, W.B. Enhancing Convolutional Neural Network Performance Using Domain Transforms in Constrained Networks. Circuits Syst Signal Process 41, 5160–5182 (2022). https://doi.org/10.1007/s00034-022-02026-2
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00034-022-02026-2