<?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">nami</journal-id><journal-title-group><journal-title xml:lang="ru">Труды НАМИ</journal-title><trans-title-group xml:lang="en"><trans-title>Trudy NAMI</trans-title></trans-title-group></journal-title-group><issn pub-type="ppub">0135-3152</issn><publisher><publisher-name>ФГУП «НАМИ»</publisher-name></publisher></journal-meta><article-meta><article-id pub-id-type="doi">10.51187/0135-3152-2021-3-37-47</article-id><article-id custom-type="elpub" pub-id-type="custom">nami-196</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>АВТОМОБИЛЕСТРОЕНИЕ</subject></subj-group><subj-group subj-group-type="section-heading" xml:lang="en"><subject>AUTOMOBILE DEVELOPMENT</subject></subj-group></article-categories><title-group><article-title>Обучение нейросетевых алгоритмов автомобильного технического зрения с учётом типов ошибок распознавания</article-title><trans-title-group xml:lang="en"><trans-title>Training neural networks for automotive computer vision systems considering types of false estimations</trans-title></trans-title-group></title-group><contrib-group><contrib contrib-type="author" corresp="yes"><name-alternatives><name name-style="eastern" xml:lang="ru"><surname>Васин</surname><given-names>П. А.</given-names></name><name name-style="western" xml:lang="en"><surname>Vasin</surname><given-names>P. A.</given-names></name></name-alternatives><bio xml:lang="ru"><p>Васин Павел Александрович, аспирант, инженер-программист, отдел видеоаналитики</p><p>г. Москва 125438, Российская Федерация</p></bio><bio xml:lang="en"><p>Vasin P.A., postgraduate, programming engineer, department of computer vision analysis</p><p>Moscow 125438, Russian Federation</p></bio><email xlink:type="simple">pavel.vasin@nami.ru</email><xref ref-type="aff" rid="aff-1"/></contrib><contrib contrib-type="author" corresp="yes"><name-alternatives><name name-style="eastern" xml:lang="ru"><surname>Куликов</surname><given-names>И. А.</given-names></name><name name-style="western" xml:lang="en"><surname>Kulikov</surname><given-names>I. A.</given-names></name></name-alternatives><bio xml:lang="ru"><p>Куликов Илья Александрович, канд. техн. наук, заведующий сектором имитационного моделирования энергетических установок</p><p>г. Москва 125438, Российская Федерация</p></bio><bio xml:lang="en"><p>Kulikov I.A., PhD (Eng), head of power unit simulation sector</p><p>Moscow 125438, Russian Federation</p></bio><email xlink:type="simple">ilya.kulyakov@nami.ru</email><xref ref-type="aff" rid="aff-2"/></contrib></contrib-group><aff-alternatives id="aff-1"><aff xml:lang="ru"><institution>Центр «Интеллектуальные системы», ГНЦ РФ ФГУП «НАМИ»</institution><country>Россия</country></aff><aff xml:lang="en"><institution>Center “Intelligence systems”, Federal State Unitary Enterprise “Central Scientific Research Automobile and Automotive Engines Institute” (FSUE “NAMI”)</institution><country>Russian Federation</country></aff></aff-alternatives><aff-alternatives id="aff-2"><aff xml:lang="ru"><institution>Центр «Энергоустановки»,  ГНЦ РФ ФГУП «НАМИ»</institution><country>Россия</country></aff><aff xml:lang="en"><institution>Centre “Power units”, Federal State Unitary Enterprise “Central Scientific Research Automobile and Automotive Engines Institute” (FSUE “NAMI”)</institution><country>Russian Federation</country></aff></aff-alternatives><pub-date pub-type="collection"><year>2021</year></pub-date><pub-date pub-type="epub"><day>05</day><month>10</month><year>2021</year></pub-date><volume>0</volume><issue>3</issue><fpage>37</fpage><lpage>47</lpage><permissions><copyright-statement>Copyright &amp;#x00A9; Васин П.А., Куликов И.А., 2021</copyright-statement><copyright-year>2021</copyright-year><copyright-holder xml:lang="ru">Васин П.А., Куликов И.А.</copyright-holder><copyright-holder xml:lang="en">Vasin P.A., Kulikov I.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://nami.elpub.ru/jour/article/view/196">https://nami.elpub.ru/jour/article/view/196</self-uri><abstract><p>Введение (постановка задачи и актуальность). В статье рассматривается проблематика обучения искусственных нейронных сетей, предназначенных для анализа изображений окружающего пространства в автомобильных системах технического зрения. Известным подходом к обучению является применение целевых функций, ориентированных на повышение общего качества идентификации безотносительно типов ложных оценок, возникающих при анализе. Однако риски для безопасности движения, связанные с типом возникающих ошибок, неравнозначны и являются наибольшими в случае ложноположительных оценок.</p><p>Цель исследования – предложить усовершенствованные целевые функции, включающие коэффициенты штрафа при ложноположительных оценках, и исследовать их влияние на поведение нейронной сети при распознавании проходимого пространства на дорогах общего пользования.</p><p>Методология и методы. Предложенные целевые функции основаны на коэффициенте Сёренсена– Дайса и различаются подходом к введению штрафа на ложноположительные срабатывания. Оценка качества работы нейронных сетей, обученных с помощью этих функций, проведена с помощью трёх метрик: коэффициента Жаккара, доли ложноположительных и доли ложноотрицательных оценок. Сравнительный анализ рассматриваемых решений проводился путём расчёта процентных соотношений между значениями метрик, которые обеспечиваются исходным вариантом целевой функции и предложенными его модификациями.</p><p>Результаты работы и научная новизна. Представлены усовершенствованные целевые функции для обучения алгоритмов технического зрения с реализацией штрафов за ложноположительные срабатывания. Экспериментальное исследование обученных нейронных сетей на тестовой выборке показало, что применение данных целевых функций позволяет снизить количество случаев ложноположительного распознавания на 21%.</p><p>Практическую значимость представляет разработанный подход к обучению нейронных сетей, позволяющий повысить безопасность автоматического управления движением транспортных средств за счёт более адекватного анализа окружающего пространства системами технического зрения.</p></abstract><trans-abstract xml:lang="en"><p>Introduction (problem statement and relevance). This article deals with the problem of training artificial neural networks intended to analyze images of the surrounding space in automotive computer vision systems. The conventional training approach implies using loss functions that only improve the overall identification quality making no distinction between types of possible false predictions. However, traffic safety risks associated with different types of prediction errors are unequal being higher for false positive estimations.</p><p>The purpose of this work is to propose improved loss functions, which include penalties for false positive predictions, and to study how using these functions affects the behavior of a convolutional neural network when estimating the drivable space.</p><p>Methodology and research methods. The proposed loss functions are based on the Sørensen-Dice coefficient differing from each other in the approaches to penalizing false positive errors. The performance of the trained neural networks is evaluated using three metrics, namely, the Jaccard coefficient, False Positive Rate and False Negative Rate. The proposed solutions are compared with the conventional one by calculating the ratios of their respective metrics.</p><p>Scientific novelty and results. The improved loss functions have been proposed to train computer vision algorithms featuring penalties for false positive estimations. The experimental study of the trained neural networks using a test dataset has shown that the improved loss functions allow reducing the False Positive Rate by 21%.</p><p>The practical significance of this work is constituted by the proposed method of training neural networks that allows to increase the safety of automated driving through an improved accuracy of analyzing the surrounding space using computer vision systems.</p></trans-abstract><kwd-group xml:lang="ru"><kwd>автомобильные системы технического зрения</kwd><kwd>проходимое пространство</kwd><kwd>нейронные сети</kwd><kwd>глубокое обучение</kwd><kwd>целевые функции</kwd></kwd-group><kwd-group xml:lang="en"><kwd>automotive computer vision systems</kwd><kwd>drivable space</kwd><kwd>neural networks</kwd><kwd>deep learning</kwd><kwd>loss functions</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">Winner H., Hakuli S., Lotz F., Singer C. Handbook of Driver Assistance Systems. Basic Information, Components and Systems for Active Safety and Comfort. – Springer International Publishing Switzerland, 2016.</mixed-citation><mixed-citation xml:lang="en">Winner H., Hakuli S., Lotz F., Singer C. Handbook of Driver Assistance Systems. Basic Information, Components and Systems for Active Safety and Comfort. Springer International Publishing Switzerland, 2016.</mixed-citation></citation-alternatives></ref><ref id="cit2"><label>2</label><citation-alternatives><mixed-citation xml:lang="ru">Pearl J. Probabilistic reasoning in intelligent systems: Networks of plausible inference. – San Francisco: Morgan Kaufmann Publishers inc., 1988. – P. 1.</mixed-citation><mixed-citation xml:lang="en">Pearl J. Probabilistic reasoning in intelligent systems: Networks of plausible inference. San Francisco, Morgan Kaufmann Publishers inc., 1988, p. 1.</mixed-citation></citation-alternatives></ref><ref id="cit3"><label>3</label><citation-alternatives><mixed-citation xml:lang="ru">Duda R., Hart P. Pattern Classification and Scene Analysis. – Hoboken: John Wiley and Sons, 1973.</mixed-citation><mixed-citation xml:lang="en">Duda R., Hart P. Pattern Clas sification and Scene Analysis. Hoboken, John Wiley and Sons, 1973.</mixed-citation></citation-alternatives></ref><ref id="cit4"><label>4</label><citation-alternatives><mixed-citation xml:lang="ru">Goodfellow I., Bengio Y., Courville A. Deep Learning. – Massachusetts: MIT Press, 2016. – P. 13.</mixed-citation><mixed-citation xml:lang="en">Goodfellow I., Bengio Y., Courville A. Deep Learning. Massachusetts, MIT Press, 2016, p. 13.</mixed-citation></citation-alternatives></ref><ref id="cit5"><label>5</label><citation-alternatives><mixed-citation xml:lang="ru">Cireşan D.C., Giusti A., Gambardella L.M., Schmidhuber Jü. Deep neural networks segment neuronal membranes in electron microscopy images / Proceedings of the 25th International Conference on Neural Information Processing Systems. – 2012. – P. 2843–2851.</mixed-citation><mixed-citation xml:lang="en">Cireşan D.C., Giusti A., Gambardella L.M., Schmidhuber Jü. Deep neural networks segment neuronal membranes in electron microscopy images. Proceedings of the 25th International Conference on Neural Information Processing Systems, 2012, pp. 2843–2851.</mixed-citation></citation-alternatives></ref><ref id="cit6"><label>6</label><citation-alternatives><mixed-citation xml:lang="ru">Krizhevsky A., Sutskever I., Hinton G.E. Imagenet classification with deep convolutional neural networks // Adv. Neural Inform Process Syst. – 2012. – P. 84–90.</mixed-citation><mixed-citation xml:lang="en">Krizhevsky A., Sutskever I., Hinton G.E. Imagenet classification with deep convolutional neural networks. Adv. Neural Inform Process Syst., 2012, pp. 84–90.</mixed-citation></citation-alternatives></ref><ref id="cit7"><label>7</label><citation-alternatives><mixed-citation xml:lang="ru">Cityscapes Dataset Benchmark Suite. URL: www.cityscapes-dataset.com/benchmarks/ (дата обращения: 26.05.2021).</mixed-citation><mixed-citation xml:lang="en">Cityscapes Dataset Benchmark Suite. Available at: www.cityscapes-dataset.com/benchmarks/ (accessed 26 May 2021).</mixed-citation></citation-alternatives></ref><ref id="cit8"><label>8</label><citation-alternatives><mixed-citation xml:lang="ru">KITTI Vision Benchmark Suite. URL: www.cvlibs.net/datasets/kitti/eval_road.php (дата обращения: 26.05.2021).</mixed-citation><mixed-citation xml:lang="en">KITTI Vision Benchmark Suite. Available at: www.cvlibs.net/datasets/kitti/eval_road.php (accessed 26 May 2021).</mixed-citation></citation-alternatives></ref><ref id="cit9"><label>9</label><citation-alternatives><mixed-citation xml:lang="ru">Rumelhart D., Hinton G., Williams R. Learning representations by back-propagating errors // Nature. – 1986. – P. 533–536</mixed-citation><mixed-citation xml:lang="en">Rumelhart D., Hinton G., Williams R. Learning representations by back-propagating errors. Nature, 1986, pp. 533–536</mixed-citation></citation-alternatives></ref><ref id="cit10"><label>10</label><citation-alternatives><mixed-citation xml:lang="ru">Kingma D.P., Adam Ba J. A Method for Stochastic Optimization. URL: https://arxiv.org/abs/1412.6980 (дата обращения: 26.05.2021).</mixed-citation><mixed-citation xml:lang="en">Kingma D.P., Adam Ba J. A Method for Stochastic Optimization. Available at: https://arxiv.org/abs/1412.6980 (accessed 26 May 2021).</mixed-citation></citation-alternatives></ref><ref id="cit11"><label>11</label><citation-alternatives><mixed-citation xml:lang="ru">Ronneberger O., Fischer P., Brox T. U-Net: Convolutional Networks for Biomedical Image Segmentation. URL: https://arxiv.org/abs/1505.04597 (дата обращения: 26.05.2021).</mixed-citation><mixed-citation xml:lang="en">Ronneberger O., Fischer P., Brox T. U-Net: Convolutional Networks for Biomedical Image Segmentation. Available at: https://arxiv.org/abs/1505.04597 (accessed 26 May 2021).</mixed-citation></citation-alternatives></ref><ref id="cit12"><label>12</label><citation-alternatives><mixed-citation xml:lang="ru">Hughes C., Chandra S., Sistu G., Horgan J., Deegan B., Chennupati S., Yogamani S. Drivespace: Towards context-aware drivable area detection // Electronic Imaging. – 2019. – P. 423–428.</mixed-citation><mixed-citation xml:lang="en">Hughes C., Chandra S., Sistu G., Horgan J., Deegan B., Chennupati S., Yogamani S. Drivespace: Towards context-aware drivable area detection. Electronic Imaging, 2019, pp. 423–428.</mixed-citation></citation-alternatives></ref><ref id="cit13"><label>13</label><citation-alternatives><mixed-citation xml:lang="ru">Pizzati F., Garcia F. Enhanced free space detection in multiple lanes based on single CNN with scene identification / Proceedings of 2019 IEEE Intelligent Vehicles Symposium (IV). – 2019. – P. 2536–2541.</mixed-citation><mixed-citation xml:lang="en">Pizzati F., Garcia F. Enhanced free space detection in multiple lanes based on single CNN with scene identification. Proceedings of 2019 IEEE Intelligent Vehicles Symposium (IV), 2019, pp. 2536–2541.</mixed-citation></citation-alternatives></ref><ref id="cit14"><label>14</label><citation-alternatives><mixed-citation xml:lang="ru">Baheti B., Innani S., Gajre S., Talbar S. Eff-Unet: A novel architecture for semantic segmentation in unstructured environment / Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition. – 2020. – P. 1473–1481.</mixed-citation><mixed-citation xml:lang="en">Baheti B., Innani S., Gajre S., Talbar S. Eff-Unet: A novel architecture for semantic segmentation in unstructured environment. Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 1473–1481.</mixed-citation></citation-alternatives></ref><ref id="cit15"><label>15</label><citation-alternatives><mixed-citation xml:lang="ru">Milletari F., Navab N., Ahmadi S. V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation / 2016 Fourth International Conference on 3D Vision (3DV). – 2016. – P. 565–571.</mixed-citation><mixed-citation xml:lang="en">Milletari F., Navab N., Ahmadi S. V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation. 2016 Fourth International Conference on 3D Vision (3DV), 2016, pp. 565–571.</mixed-citation></citation-alternatives></ref><ref id="cit16"><label>16</label><citation-alternatives><mixed-citation xml:lang="ru">Paszke A., Chaurasia A., Kim S., Culurciello E. ENet: A Deep Neural Network Architecture for RealTime Semantic Segmentation. URL: https://arxiv.org/abs/1606.02147 (дата обращения: 26.05.2021).</mixed-citation><mixed-citation xml:lang="en">Paszke A., Chaurasia A., Kim S., Culurciello E. ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation. Available at: https://arxiv.org/abs/1606.02147 (accessed 26 May 2021).</mixed-citation></citation-alternatives></ref><ref id="cit17"><label>17</label><citation-alternatives><mixed-citation xml:lang="ru">He K., Zhang X., Ren S., Sun J. Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification / Proceedings of IEEE International Conference on Computer Vision. – 2015. – P. 1026–1034.</mixed-citation><mixed-citation xml:lang="en">He K., Zhang X., Ren S., Sun J. Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification. Proceedings of IEEE International Conference on Computer Vision, 2015, pp. 1026–1034.</mixed-citation></citation-alternatives></ref><ref id="cit18"><label>18</label><citation-alternatives><mixed-citation xml:lang="ru">Cordts M., Omran M., Ramos S., Rehfeld T., Enzweiler M., Benenson R. The cityscapes dataset for semantic urban scene understanding / Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. – 2016. – P. 3213–3223.</mixed-citation><mixed-citation xml:lang="en">Cordts M., Omran M., Ramos S., Rehfeld T., Enzweiler M., Benenson R. The cityscapes dataset for semantic urban scene understanding. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 3213–3223.</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>
