ntvНаучно-технический вестник информационных технологий, механики и оптикиScientific and Technical Journal of Information Technologies, Mechanics and Optics2226-14942500-0373Университет ИТМО10.17586/2226-1494-2026-26-1-104-115ntv-569Research ArticleКОМПЬЮТЕРНЫЕ СИСТЕМЫ И ИНФОРМАЦИОННЫЕ ТЕХНОЛОГИИCOMPUTER SCIENCEРазработка алгоритма формирования рекомендаций выбора моделей детектирования объектов на основе мета-признаков данных и базы знаний экспериментовDevelopment of an algorithm for formulating recommendations for the selection of object detection models based on data meta-features and an experimental knowledge basehttps://orcid.org/0000-0002-8158-7450СметанинА. А.SmetaninA. A.

Сметанин Артём Алексеевич — главный программист; инженер

Санкт-Петербург, 196084

Санкт-Петербург, 197101

sc 57221840206

Artem A. Smetanin — Chief Programmer; Engineer

Saint Petersburg, 196084

Saint Petersburg, 197101

 sc 57221840206

 

artem_smetanin@niuitmo.ruhttps://orcid.org/0000-0002-1011-9932ДухановА. В.DukhanovA. V.

Духанов Алексей Валентинович — доктор технических наук,доцент, доцент

Санкт-Петербург, 197101

sc 56218447300

Alexey V. Dukhanov — D.Sc., Associate Professor, Associate Professor

Saint Petersburg, 197101

sc 56218447300

dukhanov@itmo.ruhttps://orcid.org/0009-0008-8005-9550ГерасимчукМ. Ю.GerasimchukM. Yu.

Герасимчук Михаил Юрьевич — ведущий инженер; аспирант

Санкт-Петербург, 197101

Санкт-Петербург, 197046

sc 57919215000

Mikhail Yu. Gerasimchuk — Leading Engineer;

Saint Petersburg, 197101

Saint Petersburg, 197046

sc 57919215000

 

mishagerasimchuk@itmo.ruhttps://orcid.org/0009-0006-4534-0262ГудовЕ. В.GudovE. V.

Гудов Евгений Валерьевич — специалист, аспирант; генеральный директор

Санкт-Петербург, 196084

Москва, 119454

Evgeny V. Gudov — Specialist, PhD Studen; CEO

Saint Petersburg, 196084

Moscow, 119454

evgenes@gmail.comООО «Планетарий 1»; Университет ИТМОРоссияPlanetarium one; ITMO UniversityRussian FederationУниверситет ИТМОРоссияITMO UniversityRussian FederationУниверситет ИТМО; ООО МТС ДИДЖИТАЛРоссияITMO University; MTS DIGITAL LLCRussian FederationООО «Планетарий 1»; Академия труда и социальных отношенийРоссияPlanetarium one; Academy of Labour and Social RelationsRussian Federation202625022026261104115Copyright © Сметанин А.А., Духанов А.В., Герасимчук М.Ю., Гудов Е.В., 20262026Сметанин А.А., Духанов А.В., Герасимчук М.Ю., Гудов Е.В.Smetanin A.A., Dukhanov A.V., Gerasimchuk M.Y., Gudov E.V.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://ntv.elpub.ru/jour/article/view/569

Введение. Детектирование объектов в системах видеонаблюдения, медицинской диагностике, в автономных транспортных средствах и иных технических системах с компонентами компьютерного зрения связано с выбором лучшей модели для конкретного набора данных. Эта задача до сих пор остается трудоемкой и требует значительных вычислительных и временных ресурсов. Предлагается метод автоматизированного формирования рекомендаций по выбору модели детектирования объектов на основе анализа мета-признаков изображений и базы знаний результатов выполненных ранее научных исследований. Метод. Предложенный метод предполагает формирование векторов мета-признаков на основе набора данных (изображения и аннотации) и поиске похожих наборов данных с использованием алгоритма CatBoostClassifier (F1-score = 0,82) в собранной базе знаний. После определения аналогичных наборов данных осуществляется ранжирование моделей детектирования объектов по точности (mean Average Precision) и производительности (Frames Per Second) на основе накопленных экспериментальных данных, что позволяет рекомендовать наиболее подходящие архитектуры. Метод реализован в открытом рекомендательном фреймворке Object Detection Recommendation System. Основные результаты. Экспериментально подтверждено, что применение предложенного алгоритма сокращает время выбора модели на 40 % по сравнению с традиционными подходами, требующими ручного тестирования. Результаты проверки на 12 тестовых наборах данных продемонстрировали высокую эффективность метода в задачах с ограниченными вычислительными ресурсами и разнообразными характеристиками изображений. Обсуждение. Благодаря интеграции в Object Detection Recommendation System метод доступен специалистам, не обладающим глубокими знаниями в области машинного обучения/компьютерного зрения. Метод особенно полезен в условиях ограниченных временных и вычислительных ресурсов, когда необходимо быстро получить обоснованное начальное приближение к выбору модели. В будущем планируется расширение базы знаний, а также адаптация метода для задач семантической сегментации и трекинга объектов.

Object detection in video surveillance systems, medical diagnostics, autonomous vehicles, and other technical systems incorporating computer vision components is closely tied to selecting the optimal model for a specific dataset. This task remains labor-intensive and requires substantial computational and time resources. In this work, the authors propose a method for automated generation of recommendations for selecting an object detection model based on the analysis of image meta-features and a knowledge base of results from previous studies. The proposed method involves forming meta-feature vectors from a dataset (images and annotations) and searching for similar datasets within a collected knowledge base using the CatBoostClassifier algorithm (F1-score = 0.82). After identifying analogous datasets, object detection models are ranked by accuracy (mAP50) and performance (FPS) based on previously accumulated experimental data, enabling the recommendation of the most suitable architectures. The method is implemented in the open-source recommendation framework Object Detection Recommendation System (ODRS). Experimental evaluation confirmed that applying the proposed algorithm reduces model selection time by 40 % compared to traditional approaches that require manual testing. Validation on 12 test datasets demonstrated high efficiency of the method in tasks with limited computational resources and diverse image characteristics. Due to its integration into ODRS, the method is accessible to specialists without deep expertise in machine learning or computer vision. It is particularly useful in scenarios with constrained time and computational resources, where a quick, well-founded initial approximation for model selection is needed. Future work includes expanding the knowledge base and adapting the method for semantic segmentation and object tracking tasks.

детектирование объектовмета-признакиавтоматизация выбора моделейкомпьютерное зрениебаза знаний экспериментовmeta-algorithmsobject detectionrecommendation systemsmeta-featuresmodel selection automationcomputer visionalgorithm optimizationexperimental knowledge baseReferencesEdozie E., Shuaibu A.N., John U.K., Sadiq B.O. Comprehensive review of recent developments in visual object detection based on deep learning // Artificial Intelligence Review. 2025 V. 58. N 9. P. 277. https://doi.org/10.1007/s10462-025-11284-wEdozie E., Shuaibu A.N., John U.K., Sadiq B.O. Comprehensive review of recent developments in visual object detection based on deep learning // Artificial Intelligence Review. 2025 V. 58. N 9. P. 277. https://doi.org/10.1007/s10462-025-11284-wYudin D., Zakharenko N., Smetanin A., Filonov R., Kichik M., Kuznetsov V., et al. Hierarchical waste detection with weakly supervised segmentation in images from recycling plants // Engineering Applications of Artificial Intelligence. 2024. V. 128. P. 128. https://doi.org/10.1016/j.engappai.2023.107542Yudin D., Zakharenko N., Smetanin A., Filonov R., Kichik M., Kuznetsov V., et al. Hierarchical waste detection with weakly supervised segmentation in images from recycling plants // Engineering Applications of Artificial Intelligence. 2024. V. 128. P. 128. https://doi.org/10.1016/j.engappai.2023.107542Alyari F., Navimipour N.J. Recommender systems: a systematic review of the state of the art literature and suggestions for future research // Kybernetes. 2018. V. 47. N 5. P. 985–1017. https://doi.org/10.1108/K-06-2017-0196Alyari F., Navimipour N.J. Recommender systems: a systematic review of the state of the art literature and suggestions for future research // Kybernetes. 2018. V. 47. N 5. P. 985–1017. https://doi.org/10.1108/K-06-2017-0196Feurer M., Klein A., Eggensperger K., Springenberg J.T., Blum M., Hutter F. Efficient and robust automated machine learning // Proc. of the 29th International Conference on Neural Information Processing Systems Systems. 2015. V. 2. P. 2755–2763.Feurer M., Klein A., Eggensperger K., Springenberg J.T., Blum M., Hutter F. Efficient and robust automated machine learning // Proc. of the 29th International Conference on Neural Information Processing Systems Systems. 2015. V. 2. P. 2755–2763.Bhanuse R., Mal S. A Systematic Review: deep learning based E-Learning Recommendation System // Proc. of the International Conference on Artificial Intelligence and Smart Systems (ICAIS). 2021. P. 190–197. https://doi.org/10.1109/icais50930.2021.9395835Bhanuse R., Mal S. A Systematic Review: deep learning based E-Learning Recommendation System // Proc. of the International Conference on Artificial Intelligence and Smart Systems (ICAIS). 2021. P. 190–197. https://doi.org/10.1109/icais50930.2021.9395835Samin H., Azim T. Knowledge based recommender system for academia using machine learning: a case study on higher education landscape of Pakistan // IEEE Access. 2019. V. 7. P. 67081–67093. https://doi.org/10.1109/ACCESS.2019.2912012Samin H., Azim T. Knowledge based recommender system for academia using machine learning: a case study on higher education landscape of Pakistan // IEEE Access. 2019. V. 7. P. 67081–67093. https://doi.org/10.1109/ACCESS.2019.2912012Rivolli A., Garcia L.P.F., Soares C., Vanschoren J., de Carvalho A.C.P.L.F. Characterizing classification datasets: a study of meta-features for meta-learning // arXiv. 2018. arXiv:1808.10406v2. https://doi.org/10.48550/arXiv.1808.10406Rivolli A., Garcia L.P.F., Soares C., Vanschoren J., de Carvalho A.C.P.L.F. Characterizing classification datasets: a study of meta-features for meta-learning // arXiv. 2018. arXiv:1808.10406v2. https://doi.org/10.48550/arXiv.1808.10406Okezie Ben-John. Beyond Accuracy: Understanding Precision and Recall in Machine Learning. [Электронный ресурс]. URL: https://benjohnokezie.medium.com/beyond-accuracy-understandingprecision-and-recall-in-machine-learning-9a07db9bc46c (дата обращения: 20.07.2025)Okezie Ben-John. Beyond Accuracy: Understanding Precision and Recall in Machine Learning. [Электронный ресурс]. URL: https://benjohnokezie.medium.com/beyond-accuracy-understandingprecision-and-recall-in-machine-learning-9a07db9bc46c (дата обращения: 20.07.2025)Gao X., Luo H., Wang Q., Zhao F., Ye L., Zhang Y. A human activity recognition algorithm based on stacking denoising autoencoder and lightGBM // Sensors. 2019. V. 19. N 4. P. 947. https://doi.org/10.3390/s19040947Gao X., Luo H., Wang Q., Zhao F., Ye L., Zhang Y. A human activity recognition algorithm based on stacking denoising autoencoder and lightGBM // Sensors. 2019. V. 19. N 4. P. 947. https://doi.org/10.3390/s19040947Salehin I., Islam M.S., Saha P., Noman S.M., Tuni A., Hasan M.M., Baten M.A. AutoML: A systematic review on automated machine learning with neural architecture search // Journal of Information and Intelligence. 2024. V. 2. N 1. P. 52–81. https://doi.org/10.1016/j.jiixd.2023.10.002Salehin I., Islam M.S., Saha P., Noman S.M., Tuni A., Hasan M.M., Baten M.A. AutoML: A systematic review on automated machine learning with neural architecture search // Journal of Information and Intelligence. 2024. V. 2. N 1. P. 52–81. https://doi.org/10.1016/j.jiixd.2023.10.002Wu B.K. K-means clustering algorithm and Python implementation // Proc. of the IEEE International Conference on Computer Science, Artificial Intelligence and Electronic Engineering (CSAIEE). 2021. P. 55–59. https://doi.org/10.1109/csaiee54046.2021.9543260Wu B.K. K-means clustering algorithm and Python implementation // Proc. of the IEEE International Conference on Computer Science, Artificial Intelligence and Electronic Engineering (CSAIEE). 2021. P. 55–59. https://doi.org/10.1109/csaiee54046.2021.9543260Manoju T., Sajeev G.P. Conductivity based agglomerative spectral clustering for community detection // Proc. of the 6th International Conference on Image Information Processing (ICIIP). 2021. P. 385–389. https://doi.org/10.1109/iciip53038.2021.9702554Manoju T., Sajeev G.P. Conductivity based agglomerative spectral clustering for community detection // Proc. of the 6th International Conference on Image Information Processing (ICIIP). 2021. P. 385–389. https://doi.org/10.1109/iciip53038.2021.9702554Dorogush A.V., Ershov V., Gulin A. CatBoost: gradient boosting with categorical features support // arXiv. 2018. arXiv:1810.11363v1. https://doi.org/10.48550/arXiv.1810.11363Dorogush A.V., Ershov V., Gulin A. CatBoost: gradient boosting with categorical features support // arXiv. 2018. arXiv:1810.11363v1. https://doi.org/10.48550/arXiv.1810.11363Mollas I., Bassiliades N., Tsoumakas G. Conclusive local interpretation rules for random forests // Data Mining and Knowledge Discovery. 2022. V. 36. N 4. P. 1521–1574. https://doi.org/10.1007/s10618-022-00839-yMollas I., Bassiliades N., Tsoumakas G. Conclusive local interpretation rules for random forests // Data Mining and Knowledge Discovery. 2022. V. 36. N 4. P. 1521–1574. https://doi.org/10.1007/s10618-022-00839-yChelmis C., Qi W. Hierarchical MultiClass AdaBoost // Proc. of the IEEE International Conference on Big Data. 2021. P. 5063–5070. https://doi.org/10.1109/bigdata52589.2021.9671291Chelmis C., Qi W. Hierarchical MultiClass AdaBoost // Proc. of the IEEE International Conference on Big Data. 2021. P. 5063–5070. https://doi.org/10.1109/bigdata52589.2021.9671291Singh G., Sachan M. Multi-layer perceptron (MLP) neural network technique for offline handwritten Gurmukhi character recognition // Proc. of the IEEE International Conference on Computational Intelligence and Computing Research. 2014. P. 1–5. https://doi.org/10.1109/iccic.2014.7238334Singh G., Sachan M. Multi-layer perceptron (MLP) neural network technique for offline handwritten Gurmukhi character recognition // Proc. of the IEEE International Conference on Computational Intelligence and Computing Research. 2014. P. 1–5. https://doi.org/10.1109/iccic.2014.7238334Alkentar S.M., Alsahwa B., Assalem A., Karakolla D. Practical comparation of the accuracy and speed of YOLO, SSD and Faster RCNN for drone detection // Journal of Engineering. 2021. V. 27. N 8. P. 19–31. https://doi.org/10.31026/j.eng.2021.08.02Alkentar S.M., Alsahwa B., Assalem A., Karakolla D. Practical comparation of the accuracy and speed of YOLO, SSD and Faster RCNN for drone detection // Journal of Engineering. 2021. V. 27. N 8. P. 19–31. https://doi.org/10.31026/j.eng.2021.08.02Diwan T., Anirudh G., Tembhurne J.V. Object detection using YOLO: challenges, architectural successors, datasets and applications // Multimedia Tools and Applications. 2023. V. 82. N 6. P. 9243–9275. https://doi.org/10.1007/s11042-022-13644-yDiwan T., Anirudh G., Tembhurne J.V. Object detection using YOLO: challenges, architectural successors, datasets and applications // Multimedia Tools and Applications. 2023. V. 82. N 6. P. 9243–9275. https://doi.org/10.1007/s11042-022-13644-yChen X.Q., Zhang Q.Y., Lin M.H., Yang G.Y., He C. No-reference color image quality assessment: from entropy to perceptual quality // EURASIP Journal on Image and Video Processing. 2019. V. 2019. N 1. P. 77. https://doi.org/10.1186/s13640-019-0479-7Chen X.Q., Zhang Q.Y., Lin M.H., Yang G.Y., He C. No-reference color image quality assessment: from entropy to perceptual quality // EURASIP Journal on Image and Video Processing. 2019. V. 2019. N 1. P. 77. https://doi.org/10.1186/s13640-019-0479-7

The authors declare that there are no conflicts of interest present.