Model of Automatic Classification and Localization of Images
https://doi.org/10.38086/2522-9613-2019-1-43-48
Abstract
The work is devoted to the identification of images in pictures, which is performed as a result of the classification and localization procedures. Analysis of models, methods and algorithms has shown that for solving the set task it is preferable to use machine learning, an artificial neural network and a genetic algorithm. The architecture of a convolutional artificial neural network is proposed. It can solve both the problem of classification and the problem of localizing images. First the network is trained, then a class is determined for the image fed to its input. Objects are localized in the image at the final stage of operations of the convolutional neural network. For this, the output values of the penultimate layer of the model are analyzed, after which the layers are traversed in the reverse order. Its goal is to find the regions with the highest response on the source image. The combined model showed acceptable results both in classification and in localization of objects. All parameters for the network are determined automatically using a genetic algorithm. Further improvement of the proposed model results will be performed by implementing distributed computing on it.
About the Authors
L. V. Serebryanaya
Belarusian State University of Informatics and Radioelectronics
Belarus
Belarus
Candidate of Science (Technology), Associate Professor, Associate Professor of the Department "Software of Information Technologies"
6 P. Brovka Str., 220013 Minsk, Republic of Belarus
K. Y. Bochkarev
Belarusian State University of Informatics and Radioelectronics
Belarus
Belarus
Undergraduate Student of the Department "Information Technologies of Automated Systems"
6 P. Brovka Str., 220013 Minsk, Republic of Belarus
A. Y. Popitich
Belarusian State University of Informatics and Radioelectronics
Belarus
Belarus
Master of Technical Sciences
6 P. Brovka Str., 220013 Minsk, Republic of Belarus
References
1. Radcliffe, N. J. Genetic set recombination and its application to neural network topology optimization. Technical Report EPCC–TR–91–21. Edinburgh: University of Edinburgh, 1991. 250 p.
2. Stanley K. О. Miikkulainen R. Evolving Neural Topologies through Augmenting Topologies. Evolutionary Computation. The MIT Press, 2002, Vol. 10 (2), pp. 99–127.
3. Simonyan K., Vedaldi A. Deep inside convolutional networks: Visualising image classification models and saliency maps. International Conference on Learning Representations Workshop. Available at: https://arxiv.org/pdf/1312.6034.pdf (accessed: 16.03.2019).
4. Perez S. Apply genetic algorithm to the learning phase of a neural network. Department of Mechanical and Aerospace Engineering University of California, Irvine, 2005. Available at: https://pdfs.semanticscholar.org/cc48/1cf3f2dfa88fc5fa84cd41d7e9f7f7de4ff2.pdf (accessed: 16.03.2019).
5. Zhou. B., Lapedriza A. Learning Deep Features for Discriminative Localization. Computer Science and Artificial Intelligence Laboratory, MIT, 2014. Available at: https://arxiv.org/pdf/1310.1531.pdf (accessed: 16.03.2019).
6. Donahue J., Jia Y., Vinyals O., Hoffman J., Zhang N., Tzeng E., Darrell T. Decaf: A deep convolutional activation feature for generic visual recognition. International Conference on Machine Learning, 2014. Available at: https://web.njit.edu/~usman/courses/cs698_spring18/RCNN.pdf (accessed: 16.03.2019).
7. Girshick R., Donahue J., Darrell T., Malik J. Rich feature hierarchies for accurate object detection and semantic segmentation, CVPR, 2014. Available at: https://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neuralnetworks.pdf (accessed: 16.03.2019).
8. Krizhevsky A., Sutskever I., Hinton G. E. Imagenet classification with deep convolutional neural networks. Advances in Neural Information Processing Systems, 2012, pp. 1097-1105.
Review
For citations:
Serebryanaya L.V., Bochkarev K.Y., Popitich A.Y. Model of Automatic Classification and Localization of Images. Digital Transformation. 2019;(1):43-48. (In Russ.) https://doi.org/10.38086/2522-9613-2019-1-43-48
Views: 3943
Email this article 