<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.3 20210610//EN" "JATS-journalpublishing1-3.dtd">
<article article-type="research-article" dtd-version="1.3" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xml:lang="ru"><front><journal-meta><journal-id journal-id-type="publisher-id">intechngu</journal-id><journal-title-group><journal-title xml:lang="ru">Вестник НГУ. Серия: Информационные технологии</journal-title><trans-title-group xml:lang="en"><trans-title>Vestnik NSU. Series: Information Technologies</trans-title></trans-title-group></journal-title-group><issn pub-type="ppub">1818-7900</issn><issn pub-type="epub">2410-0420</issn><publisher><publisher-name>НГУ</publisher-name></publisher></journal-meta><article-meta><article-id pub-id-type="doi">10.25205/1818-7900-2022-20-1-28-46</article-id><article-id custom-type="elpub" pub-id-type="custom">intechngu-188</article-id><article-categories><subj-group subj-group-type="heading"><subject>Research Article</subject></subj-group><subj-group subj-group-type="section-heading" xml:lang="ru"><subject>Статьи</subject></subj-group></article-categories><title-group><article-title>Поиск оптимальных 2D моделей нейронной сети U-net для решения задачи семантической сегментации томографических изображений гидратосодержащих образцов</article-title><trans-title-group xml:lang="en"><trans-title>Search for Optimal 2D Models of the U-net Neural Network for Solving the Problem of Semantic Segmentation of Tomographic Images of Hydrate-Containing Samples</trans-title></trans-title-group></title-group><contrib-group><contrib contrib-type="author" corresp="yes"><contrib-id contrib-id-type="orcid">https://orcid.org/0000-0002-1362-2584</contrib-id><name-alternatives><name name-style="eastern" xml:lang="ru"><surname>Колесник</surname><given-names>Т. О.</given-names></name><name name-style="western" xml:lang="en"><surname>Kolesnik</surname><given-names>T. O.</given-names></name></name-alternatives><bio xml:lang="ru"><p>Татьяна Олеговна Колесник, студентка магистратуры</p><p>Новосибирск</p></bio><bio xml:lang="en"><p>Tatyana O. Kolesnik, Master’s Student</p><p>Novosibirsk</p></bio><email xlink:type="simple">t.kolesnik@g.nsu.ru</email><xref ref-type="aff" rid="aff-1"/></contrib><contrib contrib-type="author" corresp="yes"><contrib-id contrib-id-type="orcid">https://orcid.org/0000-0002-7876-6685</contrib-id><name-alternatives><name name-style="eastern" xml:lang="ru"><surname>Дучков</surname><given-names>А. А.</given-names></name><name name-style="western" xml:lang="en"><surname>Duchkov</surname><given-names>A. A.</given-names></name></name-alternatives><bio xml:lang="ru"><p>Антон Альбертович Дучков, заведующий лабораторией</p><p>Новосибирск</p></bio><bio xml:lang="en"><p>Anton A. Duchkov, Head of the Laboratory</p><p>Novosibirsk</p></bio><email xlink:type="simple">DuchkovAA@ipgg.sbras.ru</email><xref ref-type="aff" rid="aff-2"/></contrib></contrib-group><aff-alternatives id="aff-1"><aff xml:lang="ru"><institution>Новосибирский государственный университет</institution><country>Россия</country></aff><aff xml:lang="en"><institution>Novosibirsk State University</institution><country>Russian Federation</country></aff></aff-alternatives><aff-alternatives id="aff-2"><aff xml:lang="ru"><institution>Новосибирский государственный университет; Институт нефтегазовой геологии и геофизики им. А. А. Трофимука Сибирского отделения Российской академии наук</institution><country>Россия</country></aff><aff xml:lang="en"><institution>Novosibirsk State University; Trofimuk Institute of Petroleum Geology and Geophysics of the Siberian Branch of the Russian Academy of Sciences</institution><country>Russian Federation</country></aff></aff-alternatives><pub-date pub-type="collection"><year>2022</year></pub-date><pub-date pub-type="epub"><day>17</day><month>05</month><year>2022</year></pub-date><volume>20</volume><issue>1</issue><fpage>28</fpage><lpage>46</lpage><permissions><copyright-statement>Copyright &amp;#x00A9; Колесник Т.О., Дучков А.А., 2022</copyright-statement><copyright-year>2022</copyright-year><copyright-holder xml:lang="ru">Колесник Т.О., Дучков А.А.</copyright-holder><copyright-holder xml:lang="en">Kolesnik T.O., Duchkov A.A.</copyright-holder><license xml:lang="ru" license-type="creative-commons-attribution" xlink:href="https://creativecommons.org/licenses/by/4.0/" xlink:type="simple"><license-p>Данная работа распространяется под лицензией Creative Commons Attribution 4.0.</license-p></license><license xml:lang="en" license-type="creative-commons-attribution" xlink:href="https://creativecommons.org/licenses/by/4.0/" xlink:type="simple"><license-p>This work is licensed under a Creative Commons Attribution 4.0 License.</license-p></license></permissions><self-uri xlink:href="https://intechngu.elpub.ru/jour/article/view/188">https://intechngu.elpub.ru/jour/article/view/188</self-uri><abstract><p>Задача семантической сегментации изображений гидратосодержащих пород представляет собой многоклассовую классификацию пикселей каждого 2D-изображения во входном 3D объеме по классам «Гранула», «Флюид», «Гидрат». Классы гранулы и гидрата-флюида являются слабоконтрастными, поэтому для их разделения используется модель, построенная на основе архитектуры сверточной нейронной сети U-Net. Такое решение позволяет провести классификацию пикселей по классу гранулы с точностью более 90 %, в то время как стандартная классификация по пороговому значению имеет точность лишь 56 %.</p><p>В данной статье описывается процесс поиска оптимальных моделей архитектуры U-Net посредством настройки набора гиперпараметров. Учитывая ограниченное время обработки большого объема 3D томографических данных, требовалось найти реализацию сети U-Net для достижения наилучшего качества сегментации. С другой стороны, качество сегментации может быть улучшено, посредством анализа взаимосвязей между последовательно идущими изображениями в 3D объеме. Однако 3D реализация сети является очень ресурсоемкой с точки зрения обучения модели и ее работы в режиме вывода. В связи с этим также был осуществлен поиск гиперпараметров с условием достижения сопоставимого качества сегментации при существенном упрощении реализации 2D сети. В дальнейшем найденные оптимальные гиперпараметры 2D модели могут быть использованы для настройки 3D модели сегментации.</p><p>Проведя анализ моделей по критериям сходимости и результирующего качества сегментации, мы предложили 2 модели сегментации, оптимальные с точки зрения качества и сложности модели. Рассмотренные в работе подходы гипернастройки модели U-Net имеют общий характер и могут быть применены при работе с другими наборами данных, что позволит улучшить производительность вашего нейросетевого решения как в процессе обучения, так и на стадии его эксплуатации.</p></abstract><trans-abstract xml:lang="en"><p>The task of semantic segmentation of 2D-tomographic scans of hydrate-containing rocks is a multi-class classification of pixels of each input image in a set according to the classes “Granule”, “Fluid”, “Hydrate”. Now this is implemented in the form of segmentation by the “Granule” class using the convolutional architecture of the U-Net neural network and classification of pixels unclassified as “Granule” into the “Fluid” and “Hydrate” classes by the threshold value of pixel intensity.</p><p>Considering the limited processing time of a large volume of tomographic data, it is necessary to find a compromise between the complexity of the model and the quality of segmentation. On the other hand, it is also required to propose a second, simpler implementation of the network, to extend it to a 3D segmentation model.</p><p>The solution of these optimization problems is achieved by tuning the hyperparameters of the U-Net model. To determine which set of network hyperparameters is the best in a particular case, a partial search was performed over the hyperparameter grid, limited by the variables responsible for:</p><p>1) the number of trained filters in convolution operations;</p><p>2) learning the biases vector for output channels from convolutional operations;</p><p>3) choosing an algorithm to increase the resolution in the network decoder part.</p><p>This article describes the process of finding optimal models and provides an assumption about the possibilities for their improvement.</p></trans-abstract><kwd-group xml:lang="ru"><kwd>томография</kwd><kwd>газогидраты</kwd><kwd>сверточные нейронные сети</kwd><kwd>сегментация</kwd><kwd>U-Net</kwd><kwd>поиск по сетке гиперпараметров</kwd></kwd-group><kwd-group xml:lang="en"><kwd>tomography</kwd><kwd>gas hydrates</kwd><kwd>convolutional neural networks</kwd><kwd>segmentation</kwd><kwd>U-Net</kwd><kwd>hyperparameter grid search</kwd></kwd-group></article-meta></front><back><ref-list><title>References</title><ref id="cit1"><label>1</label><citation-alternatives><mixed-citation xml:lang="ru">Chong Z. R., Yang S. H. B., Babu P., Linga P., Li X.-S. Review of natural gas hydrates as an energy resource: prospects and challenges. Applied Energy, 2016, vol. 162, pp. 1633–1652. DOI 10.1016/j.apenergy.2014.12.061</mixed-citation><mixed-citation xml:lang="en">Chong Z. R., Yang S. H. B., Babu P., Linga P., Li X.-S. Review of natural gas hydrates as an energy resource: prospects and challenges. Applied Energy, 2016, vol. 162, pp. 1633–1652. DOI 10.1016/j.apenergy.2014.12.061</mixed-citation></citation-alternatives></ref><ref id="cit2"><label>2</label><citation-alternatives><mixed-citation xml:lang="ru">Makogon Y. F., Omelchenko R. Y. Commercial gas production from Messoyakha deposit in hydrate conditions. Journal of Natural Gas Science and Engineering, 2013, vol. 11, pp. 1–6. DOI 10.1016/j.jngse.2012.08.002</mixed-citation><mixed-citation xml:lang="en">Makogon Y. F., Omelchenko R. Y. Commercial gas production from Messoyakha deposit in hydrate conditions. Journal of Natural Gas Science and Engineering, 2013, vol. 11, pp. 1–6. DOI 10.1016/j.jngse.2012.08.002</mixed-citation></citation-alternatives></ref><ref id="cit3"><label>3</label><citation-alternatives><mixed-citation xml:lang="ru">Кадыров Р. И. Рентгеновская компьютерная томография в геологии: Учеб.-метод. пособие. Казань, Казан. федеральный ун-т, 2020. 37 c.</mixed-citation><mixed-citation xml:lang="en">Kadyrov R. I. Rentgenovskaya kompyuternaya tomografiya v geologii. Uchebno-metodicheskoe posobie. Kazan, Kazan Federal University, 2020, 37 p. (in Russ.)</mixed-citation></citation-alternatives></ref><ref id="cit4"><label>4</label><citation-alternatives><mixed-citation xml:lang="ru">Дробчик А. Н., Дугаров Г. А., Дучков А. А., Купер К. Э. Акустические измерения и рентгеновская томография песчаных образцов, содержащих гидрат ксенона // Геофизические технологии. 2019. № 4. С. 17–23. DOI 10.18303/2619-1563-2019-4-17</mixed-citation><mixed-citation xml:lang="en">Drobchik A. N., Dugarov G. A., Duchkov A. A., Kuper K. E. Acoustic measurements and X-ray tomography of sand samples containing xenon hydrate. Russian Journal of Geophysical Technologies, 2019, no. 4, pp. 17–23. (in Russ.) DOI 10.18303/2619-1563-2019-4-17</mixed-citation></citation-alternatives></ref><ref id="cit5"><label>5</label><citation-alternatives><mixed-citation xml:lang="ru">Nikitin V. V., Dugarov G. A., Duchkov A. A., Fokin M. I., Drobchik A. N., Shevchenko P. D., De Carlo F., Mokso R. Dynamic in-situ imaging of methane hydrate formation and self-preservation in porous media. Marine and Petroleum Geology, 2020, vol. 115. DOI 10.1016/j.marpetgeo.2020.104234</mixed-citation><mixed-citation xml:lang="en">Nikitin V. V., Dugarov G. A., Duchkov A. A., Fokin M. I., Drobchik A. N., Shevchenko P. D., De Carlo F., Mokso R. Dynamic in-situ imaging of methane hydrate formation and self-preservation in porous media. Marine and Petroleum Geology, 2020, vol. 115. DOI 10.1016/j.marpetgeo.2020.104234</mixed-citation></citation-alternatives></ref><ref id="cit6"><label>6</label><citation-alternatives><mixed-citation xml:lang="ru">Ronneberger O., Fischer P., Brox T. U-Net: Convolutional networks for biomedical image segmentation. International Conference on Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015. Lecture Notes in Computer Science, 2015, vol. 9351, pp. 234– 241. DOI 10.1007/978-3-319-24574-4_28</mixed-citation><mixed-citation xml:lang="en">Ronneberger O., Fischer P., Brox T. U-Net: Convolutional networks for biomedical image segmentation. International Conference on Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015. Lecture Notes in Computer Science, 2015, vol. 9351, pp. 234– 241. DOI 10.1007/978-3-319-24574-4_28</mixed-citation></citation-alternatives></ref><ref id="cit7"><label>7</label><citation-alternatives><mixed-citation xml:lang="ru">Luo W., Li Y., Urtasun R., Zemel R. Understanding the effective receptive eld in deep convolutional neural networks. In: Advances in Neural Information Processing Systems. Barcelona, 2016, pp. 4898–4906.</mixed-citation><mixed-citation xml:lang="en">Luo W., Li Y., Urtasun R., Zemel R. Understanding the effective receptive eld in deep convolutional neural networks. In: Advances in Neural Information Processing Systems. Barcelona, 2016, pp. 4898–4906.</mixed-citation></citation-alternatives></ref><ref id="cit8"><label>8</label><citation-alternatives><mixed-citation xml:lang="ru">Yi-de M., Qing L., Zhi-Bai Q. Automated image segmentation using improved PCNN model based on cross-entropy. Proceedings of 2004 International Symposium on Intelligent Multimedia, Video and Speech Processing, 2004, pp. 743–746. DOI 10.1109/ISIMP.2004.1434171</mixed-citation><mixed-citation xml:lang="en">Yi-de M., Qing L., Zhi-Bai Q. Automated image segmentation using improved PCNN model based on cross-entropy. Proceedings of 2004 International Symposium on Intelligent Multimedia, Video and Speech Processing, 2004, pp. 743–746. DOI 10.1109/ISIMP.2004.1434171</mixed-citation></citation-alternatives></ref><ref id="cit9"><label>9</label><citation-alternatives><mixed-citation xml:lang="ru">Rahman M. A., Wang Y. Optimizing intersection-over-union in deep neural net-works for image segmentation. International symposium on visual computing, 2016, pp. 234–244. DOI 10.1007/978-3-319-50835-1_22</mixed-citation><mixed-citation xml:lang="en">Rahman M. A., Wang Y. Optimizing intersection-over-union in deep neural net-works for image segmentation. International symposium on visual computing, 2016, pp. 234–244. DOI 10.1007/978-3-319-50835-1_22</mixed-citation></citation-alternatives></ref><ref id="cit10"><label>10</label><citation-alternatives><mixed-citation xml:lang="ru">Kingma D., Ba J. Adam: A Method for Stochastic Optimization. In: Proceedings of the 3rd International Conference on Learning Representations, 2014.</mixed-citation><mixed-citation xml:lang="en">Kingma D., Ba J. Adam: A Method for Stochastic Optimization. In: Proceedings of the 3rd International Conference on Learning Representations, 2014.</mixed-citation></citation-alternatives></ref><ref id="cit11"><label>11</label><citation-alternatives><mixed-citation xml:lang="ru">Chollet F. Deep Learning with Python Data representations for neural networks. Shelter Island, NY, Manning Publications, 2018, pp. 31–38.</mixed-citation><mixed-citation xml:lang="en">Chollet F. Deep Learning with Python Data representations for neural networks. Shelter Island, NY, Manning Publications, 2018, pp. 31–38.</mixed-citation></citation-alternatives></ref><ref id="cit12"><label>12</label><citation-alternatives><mixed-citation xml:lang="ru">Ioffe S., Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift. Proceedings of the 32nd International Conference on Machine Learning (PMLR), 2015, vol. 37, pp. 448–456.</mixed-citation><mixed-citation xml:lang="en">Ioffe S., Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift. Proceedings of the 32nd International Conference on Machine Learning (PMLR), 2015, vol. 37, pp. 448–456.</mixed-citation></citation-alternatives></ref><ref id="cit13"><label>13</label><citation-alternatives><mixed-citation xml:lang="ru">Shi W., Caballero J., Huszar F., Totz J., Aitken A.P., Bishop R., Rueckert D., Wang Z. Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 1874–1883. DOI 10.1109/cvpr.2016.207</mixed-citation><mixed-citation xml:lang="en">Shi W., Caballero J., Huszar F., Totz J., Aitken A.P., Bishop R., Rueckert D., Wang Z. Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 1874–1883. DOI 10.1109/cvpr.2016.207</mixed-citation></citation-alternatives></ref></ref-list><fn-group><fn fn-type="conflict"><p>The authors declare that there are no conflicts of interest present.</p></fn></fn-group></back></article>
