A Convolutional Neural Network Model and Software Tool for Classifying the Presence of a Medical Mask on a Human Face
A model of a convolutional neural network, a database for training a neural network, and a software tool for classifying the presence of a medical mask on a person’s face, which allows recognizing the presence of a medical mask from the transmitted image, have been developed. The structure of the ne...
Збережено в:
| Дата: | 2023 |
|---|---|
| Автори: | , , , , , |
| Формат: | Стаття |
| Мова: | English |
| Опубліковано: |
Інститут програмних систем НАН України
2023
|
| Теми: | |
| Онлайн доступ: | https://pp.isofts.kiev.ua/index.php/ojs1/article/view/568 |
| Теги: |
Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
|
| Назва журналу: | Problems in programming |
| Завантажити файл: |  |
Репозитарії
Problems in programming| id |
pp_isofts_kiev_ua-article-568 |
|---|---|
| record_format |
ojs |
| resource_txt_mv |
ppisoftskievua/61/b136308ec40c0a59761bd17752f53561.pdf |
| spelling |
pp_isofts_kiev_ua-article-5682024-04-26T21:28:48Z A Convolutional Neural Network Model and Software Tool for Classifying the Presence of a Medical Mask on a Human Face Модель згорткової нейронної мережі та програмний засіб для класифікації наявності медичної маски на обличчі людини Hryhorenko, Y.S. Shymkovysh, V.M. Kravets, P.I. Novatskyi, A.O. Shymkovysh, L.L. Doroshenko, A.Yu. convolutional neural networks; image classification; Python; tensorflow; keras; medical masks UDC 004.93’1 медичні маски; згорткові нейронні мережі; класифікація зображень; Python; tensorflow; keras УДК 004.93’1 A model of a convolutional neural network, a database for training a neural network, and a software tool for classifying the presence of a medical mask on a person’s face, which allows recognizing the presence of a medical mask from the transmitted image, have been developed. The structure of the neural network model was optimized to improve classification results. In addition, the development of the user interface was carried out. The developed application was tested on a set of random images. The resulting model demonstrated high accuracy and robustness in solving the task of classifying the presence of a medical mask on a person’s face, which allows automating measures to protect people from the spread of diseases. The implemented application meets the requirements for speed and quality of classification. Further improvement of the classification quality of CNN can be done by collecting a larger dataset and researching other CNN architectures.Problems in programming 2023; 2: 59-66 У даній статті розглядається розробка згорткової нейронної мережі, базованої на архітектурі YOLOv4, з метою визначення наявності медичної маски на обличчі людини. Зібрано та анотовано значний набір даних, що включає різні типи масок та осіб без масок, а також застосовано аугментацію даних для покращення універсальності моделі. На основі аналізу даних проведено оптимізацію та адаптацію моделі за допомогою техніки тонкої настройки.Результати навчання та тестування отриманої моделі свідчать про її високу точність класифікації зображень. Розроблений застосунок, що використовує цю модель, має зручний графічний інтерфейс та забезпечує швидкість класифікації зображень у 20-25 кадрів за секунду.У висновках статті підкреслено, що успішність розробленої моделі у вирішенні поставленої задачі дозволяє автоматизувати заходи захисту населення від поширення інфекцій. Відзначено, що подальше покращення якості класифікації згорткової нейронної мережі можна досягнути шляхом збору більшого датасету та дослідження альтернативних архітектур нейронних мереж.Problems in programming 2023; 2: 59-66 Інститут програмних систем НАН України 2023-08-04 Article Article application/pdf https://pp.isofts.kiev.ua/index.php/ojs1/article/view/568 10.15407/pp2023.02.059 PROBLEMS IN PROGRAMMING; No 2 (2023); 59-66 ПРОБЛЕМЫ ПРОГРАММИРОВАНИЯ; No 2 (2023); 59-66 ПРОБЛЕМИ ПРОГРАМУВАННЯ; No 2 (2023); 59-66 1727-4907 10.15407/pp2023.02 en https://pp.isofts.kiev.ua/index.php/ojs1/article/view/568/619 Copyright (c) 2023 PROBLEMS IN PROGRAMMING |
| institution |
Problems in programming |
| baseUrl_str |
https://pp.isofts.kiev.ua/index.php/ojs1/oai |
| datestamp_date |
2024-04-26T21:28:48Z |
| collection |
OJS |
| language |
English |
| topic |
convolutional neural networks image classification Python tensorflow keras medical masks UDC 004.93’1 |
| spellingShingle |
convolutional neural networks image classification Python tensorflow keras medical masks UDC 004.93’1 Hryhorenko, Y.S. Shymkovysh, V.M. Kravets, P.I. Novatskyi, A.O. Shymkovysh, L.L. Doroshenko, A.Yu. A Convolutional Neural Network Model and Software Tool for Classifying the Presence of a Medical Mask on a Human Face |
| topic_facet |
convolutional neural networks image classification Python tensorflow keras medical masks UDC 004.93’1 медичні маски згорткові нейронні мережі класифікація зображень Python tensorflow keras УДК 004.93’1 |
| format |
Article |
| author |
Hryhorenko, Y.S. Shymkovysh, V.M. Kravets, P.I. Novatskyi, A.O. Shymkovysh, L.L. Doroshenko, A.Yu. |
| author_facet |
Hryhorenko, Y.S. Shymkovysh, V.M. Kravets, P.I. Novatskyi, A.O. Shymkovysh, L.L. Doroshenko, A.Yu. |
| author_sort |
Hryhorenko, Y.S. |
| title |
A Convolutional Neural Network Model and Software Tool for Classifying the Presence of a Medical Mask on a Human Face |
| title_short |
A Convolutional Neural Network Model and Software Tool for Classifying the Presence of a Medical Mask on a Human Face |
| title_full |
A Convolutional Neural Network Model and Software Tool for Classifying the Presence of a Medical Mask on a Human Face |
| title_fullStr |
A Convolutional Neural Network Model and Software Tool for Classifying the Presence of a Medical Mask on a Human Face |
| title_full_unstemmed |
A Convolutional Neural Network Model and Software Tool for Classifying the Presence of a Medical Mask on a Human Face |
| title_sort |
convolutional neural network model and software tool for classifying the presence of a medical mask on a human face |
| title_alt |
Модель згорткової нейронної мережі та програмний засіб для класифікації наявності медичної маски на обличчі людини |
| description |
A model of a convolutional neural network, a database for training a neural network, and a software tool for classifying the presence of a medical mask on a person’s face, which allows recognizing the presence of a medical mask from the transmitted image, have been developed. The structure of the neural network model was optimized to improve classification results. In addition, the development of the user interface was carried out. The developed application was tested on a set of random images. The resulting model demonstrated high accuracy and robustness in solving the task of classifying the presence of a medical mask on a person’s face, which allows automating measures to protect people from the spread of diseases. The implemented application meets the requirements for speed and quality of classification. Further improvement of the classification quality of CNN can be done by collecting a larger dataset and researching other CNN architectures.Problems in programming 2023; 2: 59-66 |
| publisher |
Інститут програмних систем НАН України |
| publishDate |
2023 |
| url |
https://pp.isofts.kiev.ua/index.php/ojs1/article/view/568 |
| work_keys_str_mv |
AT hryhorenkoys aconvolutionalneuralnetworkmodelandsoftwaretoolforclassifyingthepresenceofamedicalmaskonahumanface AT shymkovyshvm aconvolutionalneuralnetworkmodelandsoftwaretoolforclassifyingthepresenceofamedicalmaskonahumanface AT kravetspi aconvolutionalneuralnetworkmodelandsoftwaretoolforclassifyingthepresenceofamedicalmaskonahumanface AT novatskyiao aconvolutionalneuralnetworkmodelandsoftwaretoolforclassifyingthepresenceofamedicalmaskonahumanface AT shymkovyshll aconvolutionalneuralnetworkmodelandsoftwaretoolforclassifyingthepresenceofamedicalmaskonahumanface AT doroshenkoayu aconvolutionalneuralnetworkmodelandsoftwaretoolforclassifyingthepresenceofamedicalmaskonahumanface AT hryhorenkoys modelʹzgortkovoínejronnoímerežítaprogramnijzasíbdlâklasifíkacíínaâvnostímedičnoímaskinaobliččílûdini AT shymkovyshvm modelʹzgortkovoínejronnoímerežítaprogramnijzasíbdlâklasifíkacíínaâvnostímedičnoímaskinaobliččílûdini AT kravetspi modelʹzgortkovoínejronnoímerežítaprogramnijzasíbdlâklasifíkacíínaâvnostímedičnoímaskinaobliččílûdini AT novatskyiao modelʹzgortkovoínejronnoímerežítaprogramnijzasíbdlâklasifíkacíínaâvnostímedičnoímaskinaobliččílûdini AT shymkovyshll modelʹzgortkovoínejronnoímerežítaprogramnijzasíbdlâklasifíkacíínaâvnostímedičnoímaskinaobliččílûdini AT doroshenkoayu modelʹzgortkovoínejronnoímerežítaprogramnijzasíbdlâklasifíkacíínaâvnostímedičnoímaskinaobliččílûdini AT hryhorenkoys convolutionalneuralnetworkmodelandsoftwaretoolforclassifyingthepresenceofamedicalmaskonahumanface AT shymkovyshvm convolutionalneuralnetworkmodelandsoftwaretoolforclassifyingthepresenceofamedicalmaskonahumanface AT kravetspi convolutionalneuralnetworkmodelandsoftwaretoolforclassifyingthepresenceofamedicalmaskonahumanface AT novatskyiao convolutionalneuralnetworkmodelandsoftwaretoolforclassifyingthepresenceofamedicalmaskonahumanface AT shymkovyshll convolutionalneuralnetworkmodelandsoftwaretoolforclassifyingthepresenceofamedicalmaskonahumanface AT doroshenkoayu convolutionalneuralnetworkmodelandsoftwaretoolforclassifyingthepresenceofamedicalmaskonahumanface |
| first_indexed |
2024-09-16T04:08:05Z |
| last_indexed |
2024-09-16T04:08:05Z |
| _version_ |
1818528202386046976 |
| fulltext |
59
Моделі та методи машинного навчання
Introduction
Masks to protect against viruses during
the COVID-19 pandemic have become part
of the edifice for people around the world.
They protect healthy people from infected
people. However, since the virus can develop
asymptomatically, the obvious solution was the
use of masks by all people who are in public
places and on the street. Proper use of a medical
mask, especially in combination with a vaccine,
minimizes the spread of COVID-19, including
its new strains. However, the problem arose,
how to monitor whether a visitor to a public
place complies with the requirement to wear
a mask? The state and businesses began hiring
workers to control the situation. This is not the
best solution for several reasons: increased risks
for the employee’s health, additional contacts
with people in public places, the human factor,
it is difficult to provide the required number of
employees. This is where modern technologies
come to the rescue. By installing a video camera
and connecting it to the server, you can track the
presence of a medical mask on people’s faces.
This is possible thanks to neural networks.
Artificial neural networks (ANNs)
are mathematical models that imitate the
functioning of biological neural networks
and are designed to solve various problems
in the field of artificial intelligence [1-3].
They are networks of interconnected artificial
neurons that process and transmit information
among themselves. ANNs can be used in a
wide range of applications, such as computer
vision, natural language processing, speech
recognition, recommender systems, and many
others [4-9].
Convolutional neural networks
(CNNs) are a special class of artificial neural
networks that use convolutional operations
in their architecture [10-13]. They were
developed specifically for the analysis of
visual data and have shown high efficiency
in solving computer vision problems.
Convolutional networks have become
popular due to their properties, such as
locality of receptive fields, limited number
of parameters, and multi-layeredness, which
help them better learn the visual features and
structure of images.
ANNs usually contain convolutional
layers, activation functions, pooling layers,
and fully connected layers. Due to their
architecture, convolutional neural networks
are able to detect visual patterns at different
UDC 004.93’1 http://doi.org/10.15407/pp2023.02.059
Ya. S. Grigorenko, V.M. Shymkovysh, P.I. Kravets, A.O. Novatskyi, L.L. Shymkovysh,
А.Yu. Doroshenko
A CONVOLUTIONAL NEURAL NETWORK MODEL
AND SOFTWARE FOR THE CLASSIFICATION OF THE
PRESENCE OF A MEDICAL MASK ON THE HUMAN
FACE
A model of a convolutional neural network, a database for training a neural network, and a software
tool for classifying the presence of a medical mask on a person’s face, which allows recognizing the
presence of a medical mask from the transmitted image, have been developed. The structure of the neural
network model was optimized to improve classification results. In addition, the development of the user
interface was carried out. The developed application was tested on a set of random images. The resulting
model demonstrated high accuracy and robustness in solving the task of classifying the presence of a
medical mask on a person’s face, which allows automating measures to protect people from the spread
of diseases. The implemented application meets the requirements for speed and quality of classification.
Further improvement of the classification quality of CNN can be done by collecting a larger dataset and
researching other CNN architectures.
Key words: convolutional neural networks, image classification, Python, tensorflow, keras, medical masks.
© Ya. S. Grigorenko, V.M. Shymkovysh, P.I. Kravets, A.O. Novatskyi, L.L. Shymkovysh, А.Yu. Doroshenko. 2023
ISSN 1727-4907. Проблеми програмування. 2023. №2
60
Моделі та методи машинного навчання
levels of abstraction, which allows them to
classify and recognize objects in images with
high accuracy. ANNs can be used in various
applications, such as face recognition, image
classification, image segmentation, video
analysis, autonomous driving of vehicles,
and many other visual tasks.
One of the key factors behind
the popularity of convolutional neural
networks is their ability to automatically
detect and learn features in data without
the need for manual feature creation. This
contributed to the development of deep
learning and accelerated the discovery of
new applications for these networks. ANNs
have become the foundation of many modern
artificial intelligence and machine learning
technologies, and their use continues to grow
in a wide range of industries and research.
1. Development of CNN
CNN consists of 6 components:
input layer, convolutional layers, activation
functions, pooling layers, Fully connected
layers, output layer. The general structure of
a convolutional neural network is shown on
fig. 1.
Input layer: the layer at which
input data (usually images) is provided for
processing by the network.
Convolutional layers: these layers
contain a set of filters (kernels) that
«collapse» the input data by performing
convolutional operations to detect features
or image features at different levels of detail.
Activation functions: after each
convolutional layer, activation functions
(such as ReLU, sigmoid, or tanh) are applied,
which add nonlinearity to the model and
allow the network to learn more complex
functions.
Pooling layers: these layers perform
operations of increasing abstraction and
reducing the size of data by selecting the
most important informational features (for
example, using max-pooling or average-
pooling).
Fully connected layers: after a
sequence of convolutional and pooling
layers, data can be passed to fully connected
layers that perform classification or
regression based on detected features. It is
worth noting that before transferring data
to fully connected layers, they are usually
expanded into a vector.
Output layer: the last fully connected
layer gives the output values, which can be
represented in the form of a vector of class
probabilities, if the problem is classification,
or in the form of numerical values, if the
problem is regression.
To develop a convolutional neural
network, it was decided to use the basic
YOLOv4 model, as it is one of the most
powerful and effective models for object
detection [15, 16]. YOLOv4 is known for its
high accuracy and speed, which allows it to
be used in real time. These factors determine
the choice of this particular model for the
development of a convolutional neural
network aimed at classifying the presence of
a medical mask on a person’s face.
The YOLOv4 architecture is based
on a combination of several important
Figure 1. The general structure of a convolutional neural network[14]
61
Моделі та методи машинного навчання
components that contribute to its efficiency
and accuracy. The main components are:
CSPDarknet53 – basic architecture to
ensure high accuracy and relative speed of
processing;
Bag of Freebies and Bag of Specials –
sets of methods and techniques that take into
account various aspects of the network and
help improve its speed and accuracy without
additional resource costs;
PANet (Path Aggregation Network)
is an information aggregation mechanism
responsible for combining features at
different levels of abstraction to improve
object localization accuracy;
SPP (Spatial Pyramid Pooling) is a
module that allows the network to take into
account contextual information at different
scales, which increases the recognition
capabilities of the network for objects of
different sizes.
To adapt the basic YOLOv4 model
to the task of classifying the presence of
a medical mask on a person’s face, the
following studies and improvements have
been made. Data collection and annotation:
a dataset consisting of photographs of
people with and without medical masks was
collected. The data were annotated to train
the models. Fine-tuning of the model: the
technique of fine-tuning was applied to adapt
the YOLOv4 model to the specifics of the
task. The learning process included setting
hyperparameters such as learning rate, batch
size, number of epochs, and regularization,
which were adjusted to achieve optimal
accuracy and learning speed. Also, the
learning process included optimization of
the architecture. In particular, the number
of output classes was changed taking into
account the needs of the task.
Also, the sizes of convolutional
filters and the sizes of kernels were revised
to match the sizes of the faces depicted in
the photographs. Evaluation of the results:
the model was tested on the test data set
and the results were evaluated using metrics
such as accuracy, precision, recall and F1-
score. Making improvements: after fine-
tuning and evaluating the model results,
several possible directions for optimization
and refinement of the model were identified.
These improvements are aimed at increasing
the accuracy and efficiency of the model in the
task of classifying the presence of a medical
mask on a person’s face. One of the directions
is the modification of convolutional layers:
the influence of the number and size of nuclei
in convolutional layers on the accuracy and
speed of model learning was analyzed. Based
on the analysis, it was decided to change some
parameters of the convolutional layers, which
helped to improve the overall performance
of the model and its adaptation to the task of
classifying the presence of a medical mask.
2. Development of a dataset for the
training of CNN
Data collection and preparation is an
important stage in the process of developing
and training a convolutional neural network
[17-19]. This section describes the data
collection process for training the YOLOv4
model adapted to classify the presence of a
medical mask on a human face.
To ensure the representativeness of the
data and the variety of scenarios, images of
people with different types of masks, as well
as without masks, were collected. Images
have been collected from various sources
such as public databases, open websites,
and by taking photos in real-world settings.
The total volume of the collected dataset
is 5 thousand images depicting people of
different ages, genders and ethnicities.
After the images were collected,
each one was annotated, which included
identifying and labeling the face region and
the corresponding class (masked or unmasked
person). Annotation was performed manually
using a specialized software tool that allows
to create rectangular frames around the face
area and assign them appropriate class labels.
The marking process is shown in fig. 2.
In addition to data collection and
preparation, data augmentation was applied
to provide a greater variety of images and
strengthen the robustness of the model
[20-22]. Some of the augmentation techniques
that were used in this study include scaling,
rotation, panning, and illumination.
Scaling – images are scaled to
different sizes, which allows the model to
learn to recognize objects of different sizes.
62
Моделі та методи машинного навчання
Rotation – images are rotated to
arbitrary angles, which allows the model
to better adapt to different orientations of
objects.
Shift – images are shifted up, down,
left or right, which helps the model learn to
recognize objects in different positions in the
image.
Horizontal display – images are
displayed horizontally, which helps the
model adapt to symmetry and asymmetry of
objects.
Change in illumination – the intensity
of illumination in the image changes, which
allows the model to better adapt to different
lighting conditions. The collected and
annotated images were divided into training,
validation and test datasets according to the
standard 70/15/15 percentile distribution
principle.
To ensure the reliability and accuracy
of the model, it was important to ensure a
balance of classes in the training data set.
This means that the number of images with
and without masks was about the same.
After data collection, annotation,
augmentation, and segmentation, additional
data preparation was performed to train the
model. In particular, the images were scaled
to the same size corresponding to the input
size of the YOLOv4 model, and the pixel
values were normalized. This helps provide
optimal conditions for model training and
ensures better utilization of computing
resources during the training process.
As a result of data collection and
preparation, the obtained dataset was
used for training, validation and testing
of the YOLOv4 model, adapted for the
classification of the presence of a medical
mask on a person’s face.
3. Description of application
implementation
The developed application for
recognizing a medical mask on a person’s
face consisted of 2 modules: a module for
providing a graphical interface; image
classification and processing module.
The purpose of the first module is to
provide a full-fledged graphical interface
consisting of the main page, on which there is
a component for displaying the video stream
from the video camera in real time with
the result of the work of the classification
module.
The image classification and
processing module is the core of the
developed application. This module consists
of three main components: loading and
preparing the image, directly classifying
the image, and overlaying the results on the
Figure 2. Labeling of the dataset
63
Моделі та методи машинного навчання
original image. The classification component
uses a trained ANN model based on YoloV4
to classify pre-processed images coming
from a video stream. The image preparation
component provides functionality that
normalizes images and performs the
necessary transformations. This module was
implemented in the Python programming
language. Tensorflow and Keras libraries
were used to work with neural networks [23-
27]. For image processing – OpenCV and
Pillow libraries [28,29]. The Numpy library
was used to work with multidimensional
arrays.
The code of the CNN model in Python
is given in Fig. 3.
Figure 3. Code of the CNN model in Python.
64
Моделі та методи машинного навчання
4. Results of work and testing
The learning outcomes of the
developed ZNM model are as follows:
1. Accuracy of image classification on
the training set: 97.3%;
2. Accuracy of image classification on
the test set: 92.5%;
3. The average training time in the era
is 27 minutes.
Results of testing the average number
of frames per second: the speed of image
classification by trained SNM is 20-25
frames per second.
The results of testing the application
satisfy the technical requirements for its
operation. An example of the test result can
be seen in fig. 4 and fig. 5.
Conclusions
In this study, a convolutional neural
network based on the YOLOv4 architecture
was developed to classify the presence of
a medical mask on a human face. The fine-
tuning technique was applied to adapt the
model to the specifics of the task, and the
architecture and learning hyperparameters
were optimized.
A large dataset with different types of
masks and without masks was collected and
annotated to successfully train and validate
the model. The use of data augmentation
provided a greater variety of images and
improved the robustness of the model.
The resulting model demonstrated
high accuracy and robustness in solving the
task of classifying the presence of a medical
mask on a person’s face, which allows
automating measures to protect people from
the spread of diseases.
The implemented application meets
the set requirements for speed and quality of
classification.
Further improvement of the
classification quality of CNN can be done by
collecting a larger dataset and researching
other CNN architectures.
References
1. Dreyfus G (2005) Neural networks: meth-
odology and applications. Springer-Ver-
lag, Berlin. https://doi.org/10.1007/3-540-
28847-3
2. Liu W, Wang Z, Liu X, Zeng N, Liu Y, Al-
saadi FE.( 2017) A survey of deep neural
network architectures and their applications.
Neurocomputing. vol. 234, pp. 11-26.
3. Pouyanfar S, Sadiq S, Yan Y, Tian H, Tao
Y, Reyes MP, Shyu ML, Chen SC, Iyengar
S. (2018) A survey on deep learning: algo-
rithms, techniques, and applications. ACM
Comput Surv (CSUR).vol. 51, no. 5, pp.
1-36.
4. Z. -Q. Zhao, P. Zheng, S. -T. Xu and X. Wu.
(2019) Object Detection With Deep Learn-
Figure 4. The result of testing
without a mask
Figure 5. The result of testing with a mask
65
Моделі та методи машинного навчання
ing: A Review. IEEE Transactions on Neural
Networks and Learning Systems, vol. 30,
no. 11, pp. 3212-3232.
https://doi.org/10.1109/TNNLS.2018.2876865
5. Shymkovych V., Telenyk S., Kravets P.
(2021) Hardware implementation of radial-
basis neural networks with Gaussian activa-
tion functions on FPGA. Neural Computing
and Applications. vol. 33, no.15, pp. 9467-
9479. https://doi.org/10.1007/s00521-021-
05706-3
6. Tian H, Chen SC, Shyu ML.(2020) Evolu-
tionary programming based deep learning
feature selection and network construction
for visual data classification. Inf Syst Front,
vol. 22, no. 5, pp. 1053-1066.
7. Bezliudnyi Y., Shymkovysh V., Dorosh-
enko A.( 2021) Convolutional neural net-
work model and software for classification
of typical pests. Prombles in programming.
vol.4, pp. 95-102. https://doi.org/10.15407/
pp2021.04.095
8. Kravets P., Nevolko P., Shymkovych V.,
Shymkovych L. (2020) Synthesis of High-
Speed Neuro-Fuzzy-Controllers Based on
FPGA. 2020 IEEE 2nd International Con-
ference on Advanced Trends in Information
Theory (ATIT). pp. 291-295. https://doi.
org/10.1109/ATIT50783.2020.9349299
9. Shymkovych, Volodymyr, Anatoliy Dorosh-
enko, Tural Mamedov, and Olena Yatsenko
(2022) Automated Design of an Artificial
Neuron for Field-Programmable Gate Ar-
rays Based on an Algebra-Algorithmic Ap-
proach. International Scientific Technical
Journal «Problems of Control and Informat-
ics» vol. 67, no. 5, pp. 61-72.
https://doi.org/10.34229/2786-6505-2022-
5-6
10. Dhillon A, Verma GK. (2020) Convolutional
neural network: a review of models, method-
ologies and applications to object detection.
Prog Artif Intell. vol.9, no. 2, pp. 85-112.
11. Khan A, Sohail A, Zahoora U, Qureshi AS.
(2020) A survey of the recent architectures
of deep convolutional neural networks. Artif
Intell Rev. vol. 53, no. 8, pp. 5455-5501
12. Alzubaidi, L., Zhang, J., Humaidi, A.J. et al.
(2021) Review of deep learning: concepts,
CNN architectures, challenges, applications,
future directions. Journal of Big Data, vol. 8,
no. 53, pp. 1-74.
https://doi.org/10.1186/s40537-021-00444-
8
13. Khan, A., Sohail, A., Zahoora, U. et al. (2020)
A survey of the recent architectures of deep
convolutional neural networks. Artificial In-
telligence Review, vol. 53, pp. 1-62. https://
doi.org/10.1007/s10462-020-09825-6
14. Tabian I, Fu H, Sharif Khodaei Z. A. (2019)
Convolutional Neural Network for Impact
Detection and Characterization of Com-
plex Composite Structures. Sensors, vol.
19, №22:4933. https://doi.org/10.3390/
s19224933
15. Redmon, J., & Farhadi, A. (2018). YOLOv3:
An Incremental Improvement. arXiv pre-
print arXiv:1804.02767.
16. Bochkovskiy, A., Wang, C. Y., & Liao, H.
Y. M. (2020). YOLOv4: Optimal Speed and
Accuracy of Object Detection. arXiv pre-
print arXiv:2004.10934.
17. Krizhevsky, A., Sutskever, I., & Hinton,
G. E. (2012). ImageNet Classification with
Deep Convolutional Neural Networks. In
Advances in neural information processing
systems (pp. 1097-1105).
18. LeCun, Y., Bengio, Y., & Hinton, G. (2015).
Deep learning. Nature, 521(7553), 436-444.
19. Goodfellow, I., Bengio, Y., & Courville, A.
(2016). Deep Learning. MIT Press.
20. He, K., Zhang, X., Ren, S., & Sun, J. (2016).
Deep Residual Learning for Image Recogni-
tion. In Proceedings of the IEEE conference
on computer vision and pattern recognition
(pp. 770-778).
21. Shorten, C., & Khoshgoftaar, T. M. (2019). A
survey on Image Data Augmentation for Deep
Learning. Journal of Big Data, 6(1), 1-48.
22. Tan, M., & Le, Q. V. (2019). EfficientNet:
Rethinking Model Scaling for Convolutional
Neural Networks. In International Confer-
ence on Machine Learning (pp. 6105-6114).
23. TensorFlow. (n.d.). TensorFlow: An end-to-
end open source machine learning platform.
Retrieved from https://www.tensorflow.org/
24. Keras. (n.d.). Keras: The Python deep learn-
ing API. Retrieved from https://keras.io/
25. Redmon, J. (n.d.). Darknet: Open source
neural networks in C. Retrieved from https://
pjreddie.com/darknet/
26. COCO Dataset. (n.d.). Common Objects in
Context. Retrieved from https://cocodataset.
org/
66
Моделі та методи машинного навчання
27. ImageNet. (n.d.). ImageNet: A large-scale
hierarchical image database. Retrieved from
http://www.image-net.org/
28. OpenCV. (n.d.). OpenCV: Open source com-
puter vision library. Retrieved from https://
opencv.org/
29. Buslaev, A., Iglovikov, V. I., Khvedchenya,
E., Parinov, A., Druzhinin, M., & Kalinin, A.
A. (n.d.). Albumentations: Fast image aug-
mentation library. Retrieved from https://
github.com/albumentations-team/albumen-
tations
Received: 28.04.2023
Про авторів:
Григоренко Ярослав Сергійович,
студент 4 курсу Національного Техніч-
ного Універсистету України «КПІ імені
Ігоря Сікорського».
Шимкович Володимир Миколайович,
кандидат технічних наук,
доцент кафедри інформаційних систем
та технологій Національного Технічного
Універсистету України «КПІ імені Ігоря
Сікорського».
Кількість наукових публікацій в
українських виданнях – понад 30.
Кількість наукових публікацій в
зарубіжних виданнях – понад 10.
Індекс Хірша – 4. https://orcid.org/0000-
0003-4014-2786
Новацький Анатолій Олександрович,
кандидат технічних наук,
доцент кафедри інформаційних систем
та технологій Національного технічного
університету України «КПІ імені Ігоря
Сікорського».
Кількість наукових публікацій в
українських виданнях – понад 30.
Кравець Петро Іванович,
кандидат технічних наук,
доцент кафедри інформаційних систем
та технологій Національного технічного
університету України «КПІ імені Ігоря
Сікорського».
Кількість наукових публікацій в
українських виданнях – понад 40. Кіль-
кість наукових публікацій в зарубіжних
виданнях – понад 10.
Індекс Хірша – 4. https://orcid.org/0000-
0003-4632-9832
Шимкович Любов Леонідівна,
асистент кафедри інформаційних систем
та технологій Національного технічного
університету України «КПІ імені Ігоря
Сікорського».
Кількість наукових публікацій в
українських виданнях – 2.
Кількість наукових публікацій в
зарубіжних виданнях – 1. https://orcid.
org/0000-0002-1291-0373
Дорошенко Анатолій Юхимович,
доктор фізико-математичних наук,
професор, завідувач відділу теорії
комп’ютерних обчислень,
професор кафедри інформаційних систем
та технологій Національного технічного
університету України «КПІ імені Ігоря
Сікорського».
Кількість наукових публікацій в україн-
ських виданнях – понад 200. Кількість
наукових публікацій в зарубіжних видан-
нях – понад 90.
Індекс Хірша – 6. http://orcid.org/0000-
0002-8435-1451
Місце роботи авторів:
Національний технічний універси-
тет України «Київський політехніч-
ний інститут імені Ігоря Сікорського»,
проспект Перемоги 37 та
Інститут програмних систем НАН
України, 03187, м. Київ-187, проспект
Академіка Глушкова, 40.
E-mail:
yarik13371337@gmail.com,
v.shymkovych@kpi.ua,
a.novatskyi@kpi.ua,
peter_kravets@yahoo.com,
L.shymkovych@gmail.com,
doroshenkoanatoliy2@gmail.com
|