Персона:
Егоров, Алексей Дмитриевич

Организационные подразделения

Организационная единица

Институт лазерных и плазменных технологий

Стратегическая цель Института ЛаПлаз – стать ведущей научной школой и ядром развития инноваций по лазерным, плазменным, радиационным и ускорительным технологиям, с уникальными образовательными программами, востребованными на российском и мировом рынке образовательных услуг.

Организационная единица

Другие подразделения НИЯУ МИФИ

Структурные подразделения НИЯУ МИФИ, не включенные в состав институтов и факультетов.

Фамилия

Егоров

Имя

Алексей Дмитриевич

Полная страница элемента

Результаты поиска

Теперь показываю 1 - 8 из 8

Только метаданные
Comparison of the parametrically optimized implementation of Viola–Jones object detection method and MtCNN
(2021) Egorov, A. D.; Idiyatullin, A. F.; Zakirov, A. D.; Егоров, Алексей Дмитриевич
© 2021 IEEE.Face detection on images is a classical computer vision task. Solution to this problem is used in a wide range of apps, mainly in those connected with providing for the elements of administration (access to the data based on a person’s face) and elements of control (perimeter safety control systems). The most spread object’s face detecting methods may be divided in two groups: classical one (common example is Viola–Jones method) and one based on a neural network (common example is a cascaded convolutional neural network). There are plenty comparisons of face detection methods that, however, do not usually consider the potential for optimization within the parameters of realization methods in question. This paper focuses on the comparison of the parametrically optimized ways of implementation of methods which are typical for each class. Optimization is being carried out in accordance with the algorithm suggested in this paper. It reveals that the optimized Viola–Jones method is inferior to MTCNN by 20–50 % in view of the quality metrics but outruns it 7–14 times when it comes to the processing speed per image. It shows that without using algorithm optimization by parameters, it is barely impossible to get a fair quality and capacity just as the average value of the quality and capacity metrics are visibly lower than the optimal ones.
Только метаданные
Selection of hyperparameters and data augmentation method for diverse backbone models mask r-Cnn
(2021) Egorov, A. D.; Reznik, M. S.; Егоров, Алексей Дмитриевич
© 2021 IEEE.Among the most difficult computer vision tasks is one of detecting object’s action. Solving that problem, it is needed to be aware of the position of the key points of a particular type of an object. Information about key points position uses to management decision making in technical systems. It is also being complicated task with the fact that training models able to detect the key points require a significant amount of complexly organized data. This paper focuses on finding a solution to the problem of detecting the position of biological object key points. That information is useful in terms of object’s actions classification as well as for tracking them. Due to the lack of data for training, a method for obtaining additional data for training is suggested (data augmentation), also various types of backbone models are tested within the R-CNN networks on differently augmented data, with different optimizers, learning rate, number of training epochs and batches. Achieved accuracy on the test sample is more than 90%. The use of backbone models of the ResNet family allowed to achieve greater accuracy of work, which was more than 93%, while the use of reference models from the MobileNet family with an accuracy of about 90% allowed to achieve a processing speed of each frame three times higher (on average) than while using backbone models of the ResNet family.
Только метаданные
EfficientNets for DeepFake Detection: Comparison of Pretrained Models
(2021) Pokroy, A. A.; Egorov, A. D.; Егоров, Алексей Дмитриевич
© 2021 IEEE.Rapid advances in media generation techniques have made the creation of AI-generated fake face videos more accessible than ever before. In order to accelerate the development of new ways to expose forged videos, Facebook created Deep Fake Detection Challenge (DFDC), which demonstrated multiple approaches to solve this problem. Analysis of top-performing solutions revealed that all winners used pre-trained EfficientNet networks, which was finetuned on videos containing face manipulations. Because of this observation, we decide to compare the performance of EfficientNets models within the task of detecting fake videos. For comparison, we use models, based on the highest-performing entrant of DFDC, entered by Selim Seferbekov, and the DFDC dataset as training data. Our experiments show that there is no strong correlation between model performance and its size. The best accuracy was achieved by B4 and B5 models.
Только метаданные
Comparison of Combinations of Data Augmentation Methods and Transfer Learning Strategies in Image Classification Used in Convolution Deep Neural Networks
(2021) Korzhebin, T. A.; Egorov, A. D.; Егоров, Алексей Дмитриевич
© 2021 IEEE.Several studies have already made a comparison of either Data Augmentation methods or Transfer learning strategies in Convolution Deep Neural Networks for Image Classification; however, comparison of combinations of Data Augmentation methods and Transfer learning strategies remains to be accomplished. Combination of Data Augmentation methods with the highest-performing results and Transfer learning strategy with the highest-performing results does not achieve top-performing results in total as well. We make the comparison of four Data Augmentation methods, the comparison of four Transfer learning strategies, used on five different image classification models and the comparison of all combinations of them. We use small dataset consists of 40 images for training and finetuning and accuracy as metric. Our research shows that the performance results of a model with combinations of methods and strategies cannot be expected from simple comparisons of parts of this combination.
Только метаданные
Модели и методы оценки качества работы алгоритмов поиска и отслеживания объектов в видеопотоке
(2016) Егоров, А. Д.; Егоров, Алексей Дмитриевич; Роганов Е.А.
Только метаданные
Some Cases of Object Tracking Parametrs Optimization in Video Stream
(2021) Kornachuk, M.; Egorov, A. D.; Егоров, Алексей Дмитриевич
© 2021 IEEE.Object tracking in video stream can be used in many ways such as face recognition in photos, creating deepfake videos, augmented reality. A significant part of the object tracking algorithms implemented work less accurately and more slowly than they could. The main reason for this problem lies in the insufficiently careful selection of tracker parameters. This article discusses methods for solving this problem based on the analysis of tracking algorithms both using ground truth and not. The article describes methods for evaluating various parameters of tracking algorithms, and deduces patterns that occur during the operation of tracking algorithms. As a result of the study, the performance and "efficiency"of various tracking algorithms depend on their parameters, and optimal sets of parameters were selected for working on different types of data under different performance conditions.
Только метаданные
Prediction of the small molecule selectivity index against influenza virus strain A/H1N1 using machine learning methods
(2025) Egorov, A. D.; Gorohov, Ya. V.; Kuznetsov, M. M.; Borisevich, S. S.; Егоров, Алексей Дмитриевич; Борисевич, София Станиславовна
Открытый доступ
Near real-time animal action recognition and classification
(2023) Egorov, A. D.; Reznik, M. S.; Егоров, Алексей Дмитриевич