Practical Aspects of Forming Training/Test Samples for Convolutional Neural Networks

Authors

  • Yu.Ya. Tomka Yuriy Fedkovich Chernivtsi National University
  • M.V. Talakh Yuriy Fedkovich Chernivtsi National University
  • V.V. Dvorzhak Yuriy Fedkovich Chernivtsi National University
  • O.G. Ushenko Yuriy Fedkovich Chernivtsi National University

DOI:

https://doi.org/10.31649/1681-7893-2022-43-1-24-35

Keywords:

Computer Vision, Convolutional Neural Network, CNN, Deep Learning, Image Classification, Image Understanding

Abstract

The most common approaches to assessing the quality of training neural networks in the context of the problem of "small training sets" are analyzed. A review of the code implementation of the most universal approaches and ways of extending training/testing samples is carried out. The logic of the work of STN-module is analyzed. It can be inserted into existing convolutional architectures, giving neural networks the ability to actively spatially transform feature maps, conditional on the feature map itself, without any extra training supervision or modification to the optimization process.

Author Biographies

Yu.Ya. Tomka, Yuriy Fedkovich Chernivtsi National University

Ph.D., assistant professor of Computer Science Department

M.V. Talakh, Yuriy Fedkovich Chernivtsi National University

Ph.D., assistant professor of Computer Science Department

V.V. Dvorzhak, Yuriy Fedkovich Chernivtsi National University

Ph.D., assistant professor of Computer Science Department

O.G. Ushenko, Yuriy Fedkovich Chernivtsi National University

D.Sc., Professor, Head of Optics and Publishing Department

References

Krizhevsky A., Imagenet classification with deep convolutional neural networks [Electronic resource] / A.Krizhevsky, l. Sutskever, G. Hinton – 2012. – Access Mode: https://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf

Szegedy C. Going deeper with convolutions [Electronic resource] / C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V. Vanhoucke, A. Rabinovich – 2015. – Access Mode: https://arxiv.org/pdf/1409.4842.pdf

Zeiler M.D. Stochastic pooling for regularization of deep convolutional neural networks [Electronic resource] / M. D. Zeiler, R. Fergus – 2015. – Access Mode: https://arxiv.org/pdf/1301.3557.pdf

Mahendran A. Understanding deep image representations by inverting them networks [Electronic resource] / A. Mahendran, A. Vedaldi – 2015. – Access Mode: https://arxiv.org/ pdf/1312.4400.pdf

Huang C. A Diffusion-Neural-Network for Learning from Small Samples/ С. Huang, С. Moraga // International Journal of Approximate Reasoning. – 2004. – V. 35 P. 137–161.

Wang J. The Effectiveness of Data Augmentation in Image Classification using Deep Learning networks [Electronic resource] / Wang J., Perez L. Vedaldi – 2017. – Access Mode: http://cs231n.stanford.edu/reports/2017/pdfs/300.pdf

Canavet O. Efficient sample mining for object detection / O. Canavet, F. Fleuret // Proceedings of the Asian Conference on Machine Learning (ACML) – 2014 – P. 48-63.

Srivastana N. A Simple Way to Prevent Neural Networks From Overlifting / N. Srivastana, G. Hinton, A. Krizhevsky та ін.]. // The Journal of Machine Learning Research. – 2014. – Vol. 15, №1. – С. 1929–1958.

Amazon Mechanical Turk (MTurk) [Electronic resource] – Access Mode: https://www.mturk.com/.

Cires¸an D. Multi-column Deep Neural Networks for Image Classification [Electronic resource] / D. Cires¸an, U. Meier, J. Schmidhuber. – 2012. – Access Mode: https://arxiv.org/pdf/1202.2745.pdf.

LeCun Y. Traffic Sign Recognition with Multi-Scale Convolutional Networks [Electronic resource] / Y. LeCun, P. Sermanet – Access Mode: http://yann.lecun.com/exdb/publis/pdf/sermanet-ijcnn-11.pdf.

German Traffic Sign Recognition Benchmark [Electronic resource] – Access Mode: http:/benchmark.ini.rub.de/?section=gtsrb&subsection=dataset

Traffic Sign Recognition with CNN: Tools for Image Preprocessing [Electronic resource]. - 2017. - Access mode: https://habrahabr.ru/company/newprolab/blog/334618/.

Jaderberg M. Spatial Transformer Networks [Electronic resource] / M.Jaderberg, K. Simonyan, A. Zisserman, K. Kavukcuoglu. – 2016. – Access Mode: https://arxiv.org/pdf/1506.02025.pdf.

Class Imbalance Problem [Electronic resource]. – 2013. – Access Mode: http://www.chioka.in/class-imbalance-problem/.

Spearman's rank correlation coefficient [Electronic resource] – Access Mode: https://en.wikipedia.org /wiki/Spearman%27s_rank_correlation_coefficient.

Zuiderveld K. Contrast Limited Adaptive Histograph Equalization / Karel Zuiderveld. // Graphic Gems IV. San Diego: Academic Press Professional. – 1994. – С. 474–485.

Danilyuk K. ConvNets Series. Image Processing: Tools of the Trade [Electronic resource] / Kiril Danilyuk. – 2017. – Access Mode: https://towardsdatascience.com/convnets-series-image-processing-tools-of-the-trade-36e168836f0c.

Traffic sign recognition with CNN: Spatial Transformer Networks [Electronic resource]. - 2017. - Access Mode: https://habrahabr.ru/company/newprolab/blog/339484/.

СS231n Convolutional Neural Networks for Visual Recognition [Electronic resource] – Access Mode: http://cs231n.github.io/classification/.

Zakka K. Deep Learning Paper Implementations: Spatial Transformer Networks [Electronic resource] / Kevin Zakka. – 2017. – Access Mode: https://kevinzakka.github.io/2017/01/18/stn-part2/.

The MNIST - Database of handwritten digits [Electronic resource] – Access Mode: http://yann.lecun.com/exdb/mnist/.

Traffic sign recognition with Torch [Electronic resource]. – 2015. – Access Mode: https://github.com/Moodstocks/gtsrb.torch.

Vanishing gradient problem [Electronic resource] – Access Mode: https://en.wikipedia.org/ wiki/Vanishing_gradient_problem.

Olexander N. Romanyuk, Sergii V. Pavlov, and etc. "A function-based approach to real-time visualization using graphics processing units", Proc. SPIE 11581, Photonics Applications in Astronomy, Communications, Industry, and High Energy Physics Experiments 2020, 115810E (14 October 2020); https://doi.org/10.1117/12.2580212.

Leonid I. Timchenko, Natalia I. Kokriatskaia, Sergii V. Pavlov, and etc. "Q-processors for real-time image processing", Proc. SPIE 11581, Photonics Applications in Astronomy, Communications, Industry, and High Energy Physics Experiments 2020, 115810F (14 October 2020); https://doi.org/10.1117/12.2580230

Downloads

Abstract views: 95

Published

2022-12-28

How to Cite

[1]
Y. Tomka, M. Talakh, V. Dvorzhak, and O. Ushenko, “Practical Aspects of Forming Training/Test Samples for Convolutional Neural Networks”, Опт-ел. інф-енерг. техн., vol. 43, no. 1, pp. 24–35, Dec. 2022.

Issue

Section

Systems Of Technical Vision And Artificial Intelligence, Image Processing And Pattern Recognition

Metrics

Downloads

Download data is not yet available.

Most read articles by the same author(s)