УДК: 004.932

A convolutional autoencoder as a generative model of images for problems of distinguishing attributes and restoring images in missing regions

Shcherbakov, O.V., Zhdanov, I.N., Lushin, Y.A.

Publication in Journal of Optical Technology

For Russian citation (Opticheskii Zhurnal):

Щербаков О.В., Жданов И.Н., Лушин Я.А. Сверточный автоэнкодер как генеративная модель изображений для задач выделения признаков и восстановления изображений в утерянных областях // Оптический журнал. 2015. Т. 82. № 8. С. 48–53.

Shcherbakov O.V., Zhdanov I.N., Lushin Ya.A. A convolutional autoencoder as a generative model of images for problems of distinguishing attributes and restoring images in missing regions [in Russian] // Opticheskii Zhurnal. 2015. V. 82. № 8. P. 48–53.

For citation (Journal of Optical Technology):

O. V. Shcherbakov, I. N. Zhdanov, and Ya. A. Lushin, "A convolutional autoencoder as a generative model of images for problems of distinguishing attributes and restoring images in missing regions," Journal of Optical Technology. 82(8), 528-532 (2015). https://doi.org/10.1364/JOT.82.000528

Abstract:

This paper discusses an approach to the description of the structure of models capable of being trained to recognize representations of items of generative models—in particular, the architecture of a convolutional autoencoder is considered in detail. Reliable qualitative results of the operation of a convolutional encoder are also presented that show that it is valid to regard this model as generative because it is possible to implement output and sampling procedures, using as an example the solution of the problem of restoring images in missing regions.

Keywords:

generative models, convolutional autoencoder, restoring images in missing regions

Acknowledgements:

This work was carried out with the support of the Ministry of Education and Science of the Russian Federation and with the partial state support of the leading universities of the Russian Federation (Subsidy 074-U01).

OCIS codes: 150.1135

References:

1. A. S. Potapov, “Comparative analysis of structural representations of images, based on the principle of representational minimum description length,” J. Opt. Technol. 75, 715 (2008) [Opt. Zh. 75, No. 11, 35 (2008)]
2. A. S. Potapov, I. A. Malyshev, A. E. Puysha, and A. N. Averkin, “New paradigm of learnable computer vision algorithms based on the representational MDL principle,” Proc. SPIE 7696, 769606 (2010).
3. P. Vincent, H. Larochelle, I. Lajole, Y. Bengio, and P.-A. Manzagol, “Stacked denoising autoencoders: learning useful representations in a deep network with a local denoising criterion,” J. Mach. Learn. Res. 11, 3371 (2010).
4. J. Masci, U. Meier, D. Ciresan, and J. Schmidhuber, “Stacked convolutional auto-encoders for hierarchical feature extraction,” Lect. Notes Comp. Sci. 6791, 52 (2011).
5. D. Erhan, Y. Bengio, A. Courville, P.-A. Manzagol, P. Vincent, and S. Bengio, “Why does unsupervised pre-training help deep learning?” J. Mach. Learn. Res. 11, 625 (2010).
6. O. Shcherbakov and V. Batishcheva, “Image inpainting based on stacked autoencoders,” J. Phys. Conf. Ser. 536, 012020 (2014).
7. A. S. Potapov, O. V. Scherbakov, and I. N. Zhdanov, “Method of prediction based on algorithmic probability in the problem of image restoration in missing regions,” J. Opt. Technol. 80, 680 (2013) [Opt. Zh. 80, No. 11, 48 (2013)].
8. A. Potapov, O. Scherbakov, and I. Zhdanov, “Practical algorithmic probability: an image inpainting example,” Proc. SPIE 9067, 906719 (2014).