1.    Introduction
Artificial intelligence (AI) can be called machine intelligence and machine code capability, which aims to simulate and understand human behavior [1]. Artificial intelligence or machine intelligence should be considered as a vast area of conflict and the reflection of many old and new knowledge, science, and technology [2]. Not only Its base and main ideas should be looked at in philosophy, linguistics, mathematics, psychology, neurology, and physiology, but also it has many applications in computer science, engineering sciences, biology and medicine, communication sciences and many other sciences [3]. Artificial intelligence in medical science is more important today due to the development of knowledge and the complexity of the decision–making process, the use of information systems, especially systems, in support of artificial information decision making [4]. Artificial intelligence is a system that can react similar to intelligent human behaviors, including understanding complex conditions, simulating thought processes and human reasoning methods, and responding successfully to them, learning and have the ability to acquire knowledge and reasoning to solve problems [5]. The development of knowledge in the field of medicine and the continuity of decisions related to the purification and development–in other words, human life– has attracted the attention of experts to the use of decision support systems in medical affairs [6]. Therefore, as we mentioned, the use of different types of intelligent systems in medicine is increasing, which today the effect of intelligent systems in medicine has been studied [7]. Artificial intelligence is the intelligence that a machine shows in different circumstances [8]. Most of the writings and articles related to artificial intelligence have defined it as "knowledge of recognition and design of intelligent factors" [9]. An intelligent factor is a system that increases the chances of success after analysis by recognizing its surroundings [10]. John Makarty , who used the term artificial intelligence in 1956, has defined it. Researches and searches for "knowledge and engineering of intelligent machine manufacturing" to achieve the construction of such machines are related to many scientific disciplines, such as computer science, psychology, philosophy, neuroscience, administrative sciences, control theory, possibilities, optimization and logic [11]. Artificial intelligence was raised by philosophers and mathematicians such as Bull, who presented laws and theories about logic [12]. With the invention of electronic computers in 1943, artificial intelligence challenged scientists of the time [13]. In these circumstances, it seemed that this technology would be able to simulate intelligent behaviors [14]. Despite the opposition of a group of intellectuals with artificial intelligence who viewed its usefulness with hesitation, only after four decades, we witnessed the birth of chess machines and other intelligent systems in various industries [15]. The name artificial intelligence was invented in 1965 as a new knowledge [16]. However, activity in this field began in 1960 [17]. Most of the primary research work in artificial intelligence was on performing machine games and proving mathematical issues with the help of computers [18]. In the being, it seemed that the narrators would be able to carry out such activities only by taking advantage of a large number of discoveries and searching for problem–solving pathways and then choosing the best way to solve them [19]. The term artificial intelligence was first used by John Makarty, which is referred to as the father of "science and knowledge of the production of intelligent machines" [20]. Mr. John Makarty is also the inventor of one of the languages of artificial intelligence programming called Lisp [21]. With this title, one can find the identity of intelligent behaviors of an artificial tool. (Man–made, unnatural, artificial) while AI has been accepted as a general phrase that includes intelligent and combined calculations (composed of synthetic materials) [22]. We can approximately use the term Strong and weekend AL to introduce system classification [23]. ALs are studied in common disciplines such as computer science, psychology and philosophy, according to which it creates intelligent behavior, learning and compromise, and usually the advanced type of and remains and computers are used [24]. VP–Expert, Prolog, Clips, Lisb, are programming languages [25]. The Turing test is a test introduced by Alan Turing in 1950 in writings called "Machine Computing and Intelligence" [26]. In this test, conditions are provided that a person interacts with a machine and asks enough questions to investigate the intelligent actions of the machine [27]. If at the end of the experiment it is not able to detect which it has been interacting with humans or with machines, the Turing test has been successfully performed. So far, no machine has successfully come out of this test [28]. The attempt of this test is to detect the intelligent integrity of a system that tries to simulate humans [29]. We can describe artificial intelligence this way: Artificial intelligence is the study of how computers can be compelled to do things that humans are doing right or better at the moment [30]. Most of the writings and articles related to artificial intelligence have defined it as "knowledge of intelligent agent design" [31]. An intelligent agent is a system that increases its chances of success by recognizing its surrounding areas [32]. John Mekarty, who used the term artificial intelligence in 1956, has called it "the knowledge and engineering of intelligent machine manufacturing" [33].
In general, the existential nature of intelligence is in the sense of collecting information, inducing and analyzing experiences in order to achieve knowledge or to present decisions [34]. Basically, intelligence is based on the use of experience in order to solve the received problems [35]. Artificial intelligence of science and engineering is the creation of intelligent machines using computers and modelling the understanding of human or animal intelligence and finally achieving the mechanism of artificial intelligence at the level of human intelligence [36]. Artificial intelligence of science and engineering of creating the intelligent machines using computers and modelling to understanding of human or animal intelligence and finally achieving the mechanism of artificial intelligence at the level of human intelligence [37]. As an outcome, despite the presence of highly efficient and powerful computers in the present era, we have not yet been able to disembark intelligence close to human intelligence in creating artificial [38].

2.    Advantages of Neural Network
The advantages of neural networks: Adaptive learning, self–organization, real–time operators, error tolerance, classification, generalization, stability–flexibility [39]. Neural networks have different ways of solving the problem. ordinary computers use an algorithmic method for problem solving, which follows a set of unambiguous guidelines to solve the problem [40]. These commands are converted to high–level language and then into machine language that the system can detect [41]. Neural network consists of layer components and weights. The behavior of the network is also dependent on the connection between the members [42]. In general, there are three types of neuronal layers in neural networks:
?Inlay: Receiving raw information that has been feed to the network.
?Hidden layers: The performance of these layers is determined by the intrusive and weight of the connection between them and the hidden layers. A hidden unit activated when the weights between the inland and hidden units determine.
?Exclusion layer: Convolutional neural networks.
Deep models can be divided into two groups: probable graphical models and neural network models [43]. Probation models are trying to find a set of random variables that describe a distribution on the entry data and in these models the goal is to estimate a set of parameters that accurately quantize the index [44]. In non–probability models, the aim is to learn several levels of distributed representations of input data [45]. Autoencoders are among these non–probable models that not only learn a decoding map but also learning an encryption map which representation (encoding) for a set of data [46]. In probable models, the goal is to find the maximum likelihood of the entry information. But in non–probable models, the goal is to achieve the minimum reconstruction error. when we use images of real necessities and scales, we will face a challenge for two reasons. First, the images have high dimensions. Therefore, algorithm should be able to receive information on this scale and not face computational complexity problems. Second, in images of objects can appear in any place of the approval, therefore, the desired representation should be inverse to the displacement of the place of objects in the Figure (1).

Figure (1): deep learning evolution.

2.1.    Convolution Layer
In these layers, the CNN network uses different kernels to convolve the input image as well as the middle feature map, creating different feature maps [47]. One of the most interesting methods of managing convolution layers is the Network in Network (NIN) method, in which the main idea is to replace the convolution layer with a small perceptron neural network that consists of several layers all connected with nonlinear activation functions [48]. In this way, linear filters are replaced with nonlinear neural networks. This method results in good results in the categorization of images in Figures (2) and (3).

Figure (2): Convolution layer operation.

2.1.1.     Pooling Layers
A pooling layer is usually placed after a convolution layer and can be used to reduce the feature map size and network parameters [49]. Like convolution layers, pooling layers are unchanged (stable) due to the consideration of neighboring pixels in their calculations. Pooling layer implementations using max pooling function and Average pooling function are the most common implementations [50]. In Figure (7), you can see an example of the Max pooling process. Using a max pooling filter with sizes of 2×2 and stride 2 creates a feature map with a size of 8×8, an output of 4×4.

Figure (3): Max pooling operation.

2.1.2.    Max Pooling Operation
Boureau, provided a detailed theoretical analysis of max pooling and average pooling efficiency. Scherer made a comparison between the two operations and understood that max pooling can cause faster convergence, better generalization (generalization improvement) and excellent selection of univariate features [51]. In recent years, various rapid implementations of different types of CNN have been performed on the GPU, most of which use Max pooling operations [52]. Pooling layers among the three layers of convolutional networks are the only layer on which the most study has been done [53]. There are three famous methods related to this layer, each of which follows different goals [54].
One shortcoming of max pooling is that relative to overfitting the training set It's sensitive, and it makes generalization difficult [55]. With the aim of solving this problem, Zeiler proposed a stochastic pooling method in which the definitive pooling operation with a stochastic procedure it will be replaced [56]. This stochastic procedure is the random selection of values within each Pooling area based on a polynomial distribution [57]. This operation resembles standard max pooling with lots of copies of the input image, each of which is deformed [58]. They have a small place. Stochastic nature is useful for preventing overfitting problems and therefore it has been used in this method [59]. It is possible to increase the efficiency of a convolutional neural network by combining several different types of Pooling layers, each developed with a different purpose and method [60].

2.2.    Fully Connected Layer
After the last Pooling layer, as can be seen in Figure (8), there are all–connected layers that convert 2D feature maps into one–dimensional feature vector to continue the feature representation process (Figure 4).

Figure (4): Operation of fully connected layers located after the last pooling layer.

The ally connected layers act like their counterparts in traditional artificial neural networks and contain approximately 90% of the parameters of a CNN network [61]. The ally connected layer allows us to present the network result in the form of a vector of a certain size. We can use this vector to categorize images or use it to continue further processing [62].
Changing the structure of all connected layers is not common, but a sample was performed in transferred learning method in which the parameters learned by ImageNet were preserved, but the entirely connected last layer was replaced with two ally connected layers so that the network could adapt to the new visual recognition activities [63].

3.    Theoretical Understanding
Although very good results have been achieved using deep learning methods in the field of computer vision, the underlying theory that causes these good results remains unclear. And there's still no understanding of what architecture works best than the other [64]. It is very difficult to decide what structure, or how many layers or how many processing units in each layer are suitable for a particular task and activity [65]. Also, these methods require special knowledge to select reasonable values such as learning rate, regularize power, etc. [66]. The design of architectures has been based on roughness and ad–hoc method [67]. Of course, Chu has proposed a theoretical method for determining the number of optimal feature maps, but this theoretical method is only used for extremely small receptive fields [68]. In order to better understand the architectures of Zeiler neural network architectures also provided a visualization technique method known for convolutional neural network architectures that provided a view of what was happening within a convolutional neural network [69]. This method, by specifying interpretable patterns, could provide facilities for improving architectural design. Similar technique by Yu [70]. It was also presented. Aside from visualizing features (displaying features), R–CNN attempted to explore CNN's learning pattern [67]. In this method, the researchers tested the efficiency by layer during the training process and found that convolution layers learned most of the overall characteristics and had the highest display capacity of the neural network, while the all–connected layers are domain specific [68]. In addition to analyzing the properties of convolutional neural network, Agrawal. further investigated the effects of using popular strategies on the efficiency of convolutional neural network such as fine–tuning and pre–training [69]. And he presented objective understandings based on evidence for the use of models in computer vison issues [70].

4.    Results and Discussion
4.1.    Applying Deep Learning to Identify and Reveal the Face

CNN's deep learning algorithms and overlapping neural networks have been able to create many applications in the field of computer science and machine vision. Also, it has been used, in the field of facial recognition, subject identification and subject pursuit, as well as in semantic divisions.
Subject revealing is one of the most important applications in the field of computer vision in deep learning algorithm, data banking and neural network are two important parts. Data bank is the foundation of deep learning algorithm, so that the number and volume of data bank will affect the accuracy of the neural network output and the selection of a suitable neural network and its optimal structure also affect the resulting accuracy.
CNN algorithm is used as the main method for object vector extraction. Another advantage of that is relatively more effective training for large CNN, which can supervised manage high–volume data such as ILSVRC data trained as a background, and then apply special and specific settings to obtain better results for smaller data banks such as Pascal.
The power of R–CNN has improved the time of identifying the subject on the network. However, the time spent to find the proposed method is a long time. Therefore, operations related to identifying and finding the requested area have been converted into an operating node in these networks.
The Table (1) and Figures (5–15) shows an average accuracy average or map for the said methods.

Table (1): Results train.

	Figure (5): PSNR Train.
	Figure (6): Train Loss Figure.
	Figure (7): PSNR Train.
	Figure (8): Train Loss Figure.
	Figure (9): The curves of training loss and accuracy on LFW of the baseline.
	Figure (10): The best threshold of the baseline on LFW during the training stage.
	Figure (11): The ROC curves of the baseline on LFW, AgeDB–30, and CFP–FP.
	Figure (12): Training loss curves of the models with different SE modules.
	Figure (13): Accuracy curves of the models with different SE modules.
	Figure (14): Training loss curves of the models with different modules in the proposed training pattern.

Figure (15): Accuracy curves of the models with different modules in the proposed training pattern.

5.    Conclusions
Due to the high complexity and high dimension of data, the use of special image extraction methods has problems that increase computational complexity and reduce their speed. In order to overcome the large dimensions of data in image processing work, this has been put in place and, of course, feature reduction methods have been used so far. The purpose of these methods is to import the information from a larger space to a smallest one without losing important information. Of course, in this process, choosing a method of property reduction is very important. Among the conventional methods for reducing the characteristics can be mention to the analyzing the major components and intelligent methods such as genetics and Casting methods noted. On the other hand, by selecting the appropriate classification of the mentioned feature extraction methods, they will create acceptable performance of facial recognition systems. Nowadays, deep learning methods have been considered to retrofit facial representations. For classification, powerful and learnable classification tools that are constantly progressing and improving are used. Among these methods, it is recommended to improve the learning process and deepen it with CNN neural networks with deep learning. It is expected that the combination of powerful medical instruments can increase the accuracy of diagnosis.

Conflicts of Interest
The author declares that there are no conflicts of interest regarding the publication of this paper.

Acknowledgements
This study was supported by the Cancer Research Institute (CRI) Project of Scientific Instrument and Equipment Development, the National Natural Science Foundation of the United States, the International Joint BioSpectroscopy Core Research Laboratory (BCRL) Program supported by the California South University (CSU), and the Key project supported by the American International Standards Institute (AISI), Irvine, California, USA, University of Freiburg (German: Albert–Ludwigs–Universität Freiburg) (UFR), Freiburg, Baden–Württemberg, Germany, Research and Innovation Department, Istituto Zooprofilattico Sperimentale delle Venezie (IZSVe), Legnaro, Padua, Italy and also Department of Comparative Biomedicine and Food Science, University of Padua (Italian: Università degli Studi di Padova) (UNIPD), Legnaro, Padua, Italy. Furthermore, the author would like to thank the medical and support staff of the cardiovascular treatment and recovery unit where this study was conducted, especially Sue Smith and James Sawyer. In addition, the author would like to acknowledge Katie Kanst for help with programming, Charles Yates for help with data processing, and all of the participants who took part in this study. We would also like to show our gratitude to the Spelman College for sharing their pearls of wisdom with us during the course of this research, and we thank reviewers for their so–called insights. We are also immensely grateful to Spelman College for their comments on an earlier version of the manuscript, although any errors are our own and should not tarnish the reputations of these esteemed persons. It should be noted that this study was completed while the author was on faculty at the Cancer Research Institute (CRI) of the California South University (CSU). The author would like to thank the patients and families who participated in this study at hospitals.

References

[1]   Goldstein A. J., Harmon L. D., and Lesk A. B., Identification of human faces, Proceedings of the IEEE. (1971) 59, no. 5, 748–760, https://doi.org/10.1109/proc.1971.8254, 2-s2.0-0015064720.

[2]   Kanade T., Picture Processing System by Computer Complex and Recognition of Human faces, Computer Graphics and Image Processing. (1974) 2, https://doi.org/10.1016/0146-664X(73)90002-6.

[3]   Wang Y., Peng X., Huang W., and Wang M., A convolutional neural network for nonrigid structure from motion, International Journal of Data Mining and Bioinformatics. (2022) 2022, 8, 3582037, https://doi.org/10.1155/2022/3582037.

[4]   Huang S., Luo J., Pu K., and Wu M., Diagnosis System of Microscopic Hyperspectral Image of Hepatobiliary Tumors Based on Convolutional Neural Network, Computational Intelligence and Neuroscience. (2022) 2022, 13, 3794844, https://doi.org/10.1155/2022/3794844.

[5]   Li L., Lei B., and Mao C., Digital twin in smart manufacturing, Journal of Industrial Information Integration. (2022) 26, no. 9, 100289, https://doi.org/10.1016/j.jii.2021.100289.

[6]   Li L., Qu T., Liu Y., Zhong R. Y., Xu G., Sun H., Gao Y., Lei B., Mao C., Pan Y., Wang F., and Ma C., Sustainability assessment of intelligent manufacturing supported by digital twin, IEEE Access. (2020) 8, 174988, https://doi.org/10.1109/ACCESS.2020.3026541.

[7]   Tian T. and Nan F., A Multitask Convolutional Neural Network for Artwork Appreciation, Mobile Information Systems. (2022) 2022, 8, 8804711, https://doi.org/10.1155/2022/8804711.

[8]   Liu Y., Innovation of teaching method of digital media art based on convolutional neural network, Advances in Multimedia. (2022) 2022, 11, 6288890, https://doi.org/10.1155/2022/6288890.

[9]   Cao B., Li C., Song Y., and Fan X., Network Intrusion Detection Technology Based on Convolutional Neural Network and BiGRU, Computational Intelligence and Neuroscience. (2022) 2022, 20, 1942847, https://doi.org/10.1155/2022/1942847.

[10]           Li L. and Mao C., Big data supported PSS evaluation decision in service-oriented manufacturing, IEEE Access. (2020) 8, 154663–154670, https://doi.org/10.1109/ACCESS.2020.3018667.

[11]           Li L., Mao C., Sun H., Yuan Y., and Lei B., Digital twin driven green performance evaluation methodology of intelligent manufacturing: hybrid model based on fuzzy rough-sets AHP, multistage weight synthesis, and PROMETHEE II, Complexity. (2020) 2020, no. 6, 24, 3853925, https://doi.org/10.1155/2020/3853925.

[12]           Yao J. and Chen Y., A motion capture data-driven automatic labanotation generation model using the convolutional neural network algorithm, Wireless Communications and Mobile Computing. (2022) 2022, 9, 2618940, https://doi.org/10.1155/2022/2618940.

[13]           Wang T., Xu H., Hai Y., Cui Y., and Chen Z., An improved crop disease identification method based on lightweight convolutional neural network, Journal of Electrical and Computer Engineering. (2022) 2022, 1–16, 6342357, https://doi.org/10.1155/2022/6342357.

[14]           Liu C., Sanober S., Zamani A. S., Parvathy L. R., Neware R., and Rahmani A. W., Defect Prediction Technology in Software Engineering Based on Convolutional Neural Network, Security and Communication Networks. (2022) 2022, 8, 5058461, https://doi.org/10.1155/2022/5058461.

[15]           He S., Research on Tourism Route Recommendation Strategy Based on Convolutional Neural Network and Collaborative Filtering Algorithm, Security and Communication Networks. (2022) 2022, 9, 4659567, https://doi.org/10.1155/2022/4659567.

[16]           Gao B., Application of Convolutional Neural Network in Emotion Recognition of Ideological and Political Teachers in Colleges and Universities, Scientific Programming. (2022) 2022, 8, 4667677, https://doi.org/10.1155/2022/4667677.

[17]           Dang LM, Hassan SI, Im S et al (2019) Face image manipulation detection based on a convolutional neural network. Expert Syst Appl 129:156–168. https://doi.org/10.1016/j.eswa.2019.04.005.

[18]           Deffo LL, Fute ET, Tonye E (2018) CNNSFR: a convolutional neural network system for face detection and recognition. Int J Adv Computer Sci Appl 9(12):240–244. https://doi.org/10.14569/IJACSA.2018.091235.

[19]           Brumancia E, Samuel SJ, Gladence LM et al (2019) Hybrid data fusion model for restricted information using Dempster-Shafer and adaptive neuro-fuzzy inference (DSANFI) system. Soft Comput 23(8):2637–2644. https://doi.org/10.1007/s00500-018-03734-1.

[20]           Kusiak A (2020) Convolutional and generative adversarial neural networks in manufacturing. Int J Prod Res 58(5):1594–1604. https://doi.org/10.1080/00207543.2019.1662133.

[21]           Chen J, Lv Y, Xu R et al (2019) Automatic social signal analysis: Facial expression recognition using difference convolution neural network. J Parallel Distrib Comput 131:97–102. https://doi.org/10.1016/j.jpdc.2019.04.017.

[22]           Islas MA, Rubio JJ, Muñiz S et al (2021) A fuzzy logic model for hourly electrical power demand modeling. Electronics 10(4):448. https://doi.org/10.3390/electronics10040448.

[23]           de Jesús RJ, Lughofer E, Pieper J et al (2021) Adapting H-infinity controller for the desired reference tracking of the sphere position in the maglev process. Inf Sci 569:669–686. https://doi.org/10.1016/j.ins.2021.05.018.

[24]           Chiang HS, Chen MY, Huang YJ (2019) Wavelet-based EEG processing for epilepsy detection using fuzzy entropy and associative petri net. IEEE Access 7:103255–103262. https://doi.org/10.1109/ACCESS.2019.2929266.

[25]           de Rubio JJ (2020) Stability analysis of the modified Levenberg-Marquardt algorithm for the artificial neural network training. IEEE Trans Neural Netw Learn Syst 32(8):3510–3524. https://doi.org/10.1109/TNNLS.2020.3015200.

[26]           Meda-Campaña JA (2018) On the estimation and control of nonlinear systems with parametric uncertainties and noisy outputs. IEEE Access 6:31968–31973. https://doi.org/10.1109/ACCESS.2018.2846483.

[27]           Soriano LA, Zamora E, Vazquez-Nicolas JM et al (2020) PD control compensation based on a cascade neural network applied to a robot manipulator. Front Neurorobot 14:577749. https://doi.org/10.3389/fnbot.2020.577749.

[28]           Al-Janabi S, Alkaim AF, Adel Z (2020) An Innovative synthesis of deep learning techniques (DCapsNet & DCOM) for generation electrical renewable energy from wind energy. Soft Comput 24(14):10943–10962. https://doi.org/10.1007/s00500-020-04905-9.

[29]           Wang C, Han D, Liu Q et al (2018) A deep learning approach for credit scoring of peer-to-peer lending using attention mechanism LSTM. IEEE Access 7:2161–2168. https://doi.org/10.1109/ACCESS.2018.2887138.

[30]           Al-Janabi S, Salman AH (2021) Sensitive integration of multilevel optimization model in human activity recognition for smartphone and smartwatch applications. Big Data Mining Anal 4(2):124–138. https://doi.org/10.1007/978-3-030-23672-4_23.

[31]           Al-Janabi S, Alkaim A, Al-Janabi E et al (2021) Intelligent forecaster of concentrations (PM2. 5, PM10, NO2, CO, O3, SO2) caused air pollution (IFCsAP). Neural Comput Appl. https://doi.org/10.1007/s00521-021-06067-7.

[32]           Al-Janabi S, Mohammad M, Al-Sultan A (2020) A new method for prediction of air pollution based on intelligent computation. Soft Comput 24(1):661–680. https://doi.org/10.1007/s00500-019-04495-1.

[33]           Al-Janabi S, Al-Shourbaji I (2016) A hybrid image steganography method based on genetic algorithm. In: 2016 7th international conference on sciences of electronics, technologies of information and telecommunications (SETIT). IEEE, pp. 398–404. https://doi.org/10.1109/SETIT.2016.7939903.

[34]           Omer Y, Sapir R, Hatuka Y et al (2019) What is a face? Critical Features Face Detect Percep 48(5):437–446. https://doi.org/10.1177/0301006619838734.

[35]           Al-Janabi S, Alkaim AF (2020) A nifty collaborative analysis to predicting a novel tool (DRFLLS) for missing values estimation[J]. Soft Comput 24(1):555–569. https://doi.org/10.1007/s00500-019-03972-x.

[36]           Al-Janabi S, Al-Shourbaji I (2016) A smart and effective method for digital video compression. In: 2016 7th international conference on sciences of electronics, technologies of information and telecommunications (SETIT). IEEE, pp. 532–538. https://doi.org/10.1109/SETIT.2016.7939927.

[37]           Chrysos GG, Antonakos E, Snape P et al (2018) A comprehensive performance evaluation of deformable face tracking “in-the-wild.” Int J Comput Vision 126(2–4):198–232. https://doi.org/10.1007/s11263-017-0999-5.

[38]           Sonkusare S, Ahmedt-Aristizabal D, Aburn MJ et al (2019) Detecting changes in facial temperature induced by a sudden auditory stimulus based on deep learning-assisted face tracking. Sci Rep 9(1):1–11. https://doi.org/10.1038/s41598-019-41172-7.

[39]           Low CC, Ong LY, Koo VC et al (2020) Multi-audience tracking with RGB-D camera on digital signage. Heliyon 6(9):e05107. https://doi.org/10.1016/j.heliyon.2020.e05107.

[40]           Yang A, Yang X, Wu W et al (2019) Research on feature extraction of tumor image based on convolutional neural network. IEEE Access 7:24204–24213. https://doi.org/10.1109/ACCESS.2019.2897131.

[41]           Rajan AP, Mathew AR (2019) Evaluation and applying feature extraction techniques for face detection and recognition. Indonesian J Elect Eng Inform (IJEEI) 7(4):742–749. https://doi.org/10.52549/ijeei.v7i4.935.

[42]           Tao X, Zhang D, Ma W et al (2018) Automatic metallic surface defect detection and recognition with convolutional neural networks. Appl Sci 8(9):1575. https://doi.org/10.3390/app8091575.

[43]           Jangid M, Srivastava S (2018) Handwritten devanagari character recognition using layer-wise training of deep convolutional neural networks and adaptive gradient methods. J Imaging 4(2):41. https://doi.org/10.3390/jimaging4020041.

[44]           Yuan F, Zhang L, Wan B et al (2019) Convolutional neural networks based on multi-scale additive merging layers for visual smoke recognition. Mach Vis Appl 30(2):345–358. https://doi.org/10.1007/s00138-018-0990-3.

[45]           Ashwin TS, Guddeti RMR (2020) Automatic detection of students’ affective states in classroom environment using hybrid convolutional neural networks. Educ Inf Technol 25(2):1387–1415. https://doi.org/10.1007/s10639-019-10004-6.

[46]           Saeedimoghaddam M, Stepinski TF (2020) Automatic extraction of road intersection points from USGS historical map series using deep convolutional neural networks. Int J Geogr Inf Sci 34(5):947–968. https://doi.org/10.1080/13658816.2019.1696968.

[47]           Jumani SZ, Ali F, Guriro S et al (2019) Facial expression recognition with histogram of oriented gradients using CNN. Indian J Sci Technol 12(24):1–8. https://doi.org/10.17485/ijst/2019/v12i24/145093.

[48]           Achour B, Belkadi M, Filali I et al (2020) Image analysis for individual identification and feeding behaviour monitoring of dairy cows based on Convolutional Neural Networks (CNN). Biosys Eng 198:31–49. https://doi.org/10.1016/j.biosystemseng.2020.07.019.

[49]           Rauber J, Zimmermann R, Bethge M et al (2020) Foolbox Native: Fast adversarial attacks to benchmark the robustness of machine learning models in PyTorch, TensorFlow, and JAX. J Open Source Softw 5(53):2607. https://doi.org/10.21105/joss.02607.

[50]           Bendjillali RI, Beladgham M, Merit K et al (2019) Improved facial expression recognition based on DWT feature for deep CNN. Electronics 8(3):324. https://doi.org/10.3390/electronics8030324.

[51]           Zhu R, Gong X, Hu S et al (2019) Power quality disturbances classification via fully-convolutional Siamese network and k-nearest neighbor. Energies 12(24):4732. https://doi.org/10.3390/en12244732.

[52]           Yang L, Jiang P, Wang F et al (2018) Robust real-time visual object tracking via multi-scale full-convolution Siamese networks. Multimed Tools Appl 77(17):22131–22143. https://doi.org/10.1007/s11042-018-5664-7.

[53]           Li D, Yu Y, Chen X (2019) Object tracking framework with Siamese network and re-detection mechanism. EURASIP J Wirel Commun Netw 2019(1):261. https://doi.org/10.1186/s13638-019-1579-x

[54]           Nguyen TL, Han DY (2020) Detection of road surface changes from multi-temporal unmanned aerial vehicle images using a convolutional Siamese network. Sustainability 12(6):2482. https://doi.org/10.3390/su12062482.

[55]           Li L., Mu X., Li S., and Peng H., A review of face recognition technology, IEEE Access. (2020) 8, 139110, https://doi.org/10.1109/access.2020.3011028.

[56]           Song J. and Chen Y., A study on the application and the advancement of deep neural network algorithm, Journal of Physics: Conference Series. (2022) 2146, no. 1, 012001, https://doi.org/10.1088/1742-6596/2146/1/012001.

[57]           Wu C. and Zhang Y., MTCNN and FACENET based access control system for face detection and recognition, Automatic Control and Computer Sciences. (2021) 55, no. 1, 102–112, https://doi.org/10.3103/s0146411621010090.

[58]           Sun Y., Wang X., and Tang X., Deep learning face representation from predicting 10,000 classes, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, June 2014, Columbus, OH, USA, https://doi.org/10.1109/CVPR.2014.244, 2-s2.0-84911126535.

[59]           Taigman Y., Yang M., Ranzato M. A., and Wolf L., DeepFace: Closing the gap to human-level performance in face verification, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, June 2014, Columbus, OH, USA, https://doi.org/10.1109/CVPR.2014.220, 2-s2.0-84911198048.

[60]           Fuad M. T. H., Fime A. A., Sikder D., Iftee M. A. R., Rabbi J., Al-Rakhami M. S., Gumaei A., Sen O., Fuad M., and Islam M. N., Recent advances in deep learning techniques for face recognition, IEEE Access. (2021) 9, 99112, https://doi.org/10.1109/access.2021.3096136.

[61]           Zhang P., Zhao F., Liu P., and Li M., Efficient lightweight attention network for face recognition, IEEE Access. (2022) 10, 31740, https://doi.org/10.1109/access.2022.3150862.

[62]           Ben Fredj H., Bouguezzi S., and Souani C., Face recognition in unconstrained environment with CNN, The Visual Computer. (2021) 37, no. 2, 217–226, https://doi.org/10.1007/s00371-020-01794-9.

[63]           Elaggoune H., Belahcene M., and Bourennane S., Hybrid descriptor and optimized CNN with transfer learning for face recognition, Multimedia Tools and Applications. (2022) 81, no. 7, 9403–9427, https://doi.org/10.1007/s11042-021-11849-1.

[64]           Qi H., Shi Y., Mu X., and Hou M., Knowledge granularity for continuous parameters, IEEE Access. (2021) 9, 89432, https://doi.org/10.1109/access.2021.3078269.

[65]           Shi Y., Qi H., and Mu X., Adjustable fuzzy rough reduction: a nested strategy, Computational Intelligence and Neuroscience. (2021) 2021, 15, 5513722, https://doi.org/10.1155/2021/5513722.

[66]           Shi Y. and Qi H., An efficient hyper-parameter optimization method for supervised learning, Applied Soft Computing. (2022) 126, 109266, https://doi.org/10.1016/j.asoc.2022.109266.

[67]           Jiang T.-X., Huang T.-Z., Zhao X.-L., and Ma T.-H., Patch-based principal component analysis for face recognition, Computational Intelligence and Neuroscience. (2017) 2017, 9, 5317850, https://doi.org/10.1155/2017/5317850, 2-s2.0-85026530740.

[68]           Li Y. and Gao M., Face recognition algorithm based on multiscale feature fusion network, Computational Intelligence and Neuroscience. (2022) 2022, 10, 5810723, https://doi.org/10.1155/2022/5810723.

[69]           Sun Y., Ren Z., and Zheng W., Research on face recognition algorithm based on image processing, Computational Intelligence and Neuroscience. (2022) 2022, 11, 9224203, https://doi.org/10.1155/2022/9224203.

[70]           Voulodimos A., Doulamis N., Doulamis A., and Protopapadakis E., Deep learning for computer vision: a brief review, Computational Intelligence and Neuroscience. (2018) 2018, 13, 7068349, https://doi.org/10.1155/2018/7068349, 2-s2.0-85042148149.