Computer Science and Information Technologies

Vol. 7, No. 1, March 2026, pp. 1~9

ISSN: 2722-3221, DOI: 10.11591/csit.v7i1.pp1-9  1

Implementation of face recognition using Python

Febrian Wahyu Christanto1, Husnul Arifin1, Christine Dewi2,3, Teguh Prasandy4

1Department of Information Technology, Faculty of Information Technology and Communication, Universitas Semarang, Semarang,
Indonesia
2Department of Information Technology, Faculty of Information Technology, Satya Wacana Christian University, Salatiga, Indonesia
3School of Information Technology, Faculty of Science Engineering and Built Environment, Deakin University, Melbourne, Australia
4Department of Information System, Binus On-line Learning, Bina Nusantara University, Jakarta, Indonesia

Article Info ABSTRACT

Article history: Artificial intelligence (AI)-based technology systems are developing rapidly.
Along with technological development the number of criminal cases caused
Received May 18, 2025 by facial forgery is also growing. Cases of theft and housebreaking with fake
Revised Jul 22, 2025 photos are a common problem in Semarang. In 2022–2023 the number of
Accepted Nov 28, 2025 cases of theft and housebreaking reached 372,965 with a crime risk level of
137/100,000 people. To overcome this problem the facial recognition system
used in the door security system uses digital image processing. This method
Keywords: works by imitating how nerve cells communicate with interconnected
neurons, or more precisely, how artificial neural networks function in
Artificial intelligence humans. As training data, image capture and facial recognition are carried
Digital image processing out using a webcam and the Python programming language with the
Environmental security TensorFlow library. The image processing algorithm uses 400 facial images
Face recognition with an accuracy rate of 95%. However further development is needed to
TensorFlow library improve the efficiency and accuracy of the system to produce better results.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Febrian Wahyu Christanto
Department of Information Technology, Faculty of Information Technology and Communication
Universitas Semarang
Soekarno-Hatta Arterial Road, Tlogosari, Semarang, Indonesia
Email: [Link]@[Link]

1. INTRODUCTION
Home security is important to avoid crime problems such as house robbery. According to the central
statistics agency of Semarang city, Indonesia, crime cases in 2022-2023 reached 372,965 with a risk level of
137/100,000 residents [1]. The increasing number of theft and robbery cases in housing is a consideration for
tightening the current security system. Although most housing is provided with closed-circuit television
(CCTV), facial recognition technology is needed to ensure that only people with access are allowed to enter
the house [2]. The use of password-type security such as using letters and numbers often results in errors and
is vulnerable to hacking because based on CNBC Indonesia data in 2022, the password hacking rate
increased by 921 attacks per second, plus data according to the Microsoft Digital Defense Report shows an
increase of up to 74% in one year. Therefore, the use of facial passwords is the best solution to prevent house
theft [3]. Deep learning is developed using a machine learning approach that has a working principle of a
facial detection camera system with an accuracy equivalent to the ability of the human brain. Even the
accuracy produced reaches 98.8% on a 64×64×3 face, but is also supported by the quality of the camera used
to process facial recognition. Its implementation uses the single-shot detector (SSD) method, which is one of
the digital image processing methods [4], [5].
The use of digital image processing methods is a method that has the highest accuracy both statistically
and in real time [6]. The recorded accuracy level is 100% when the viewing angle is facing forward, then when

Journal homepage: [Link]

2  ISSN: 2722-3221

not facing forward, the highest accuracy level is 80%, but with a distance range of 60-250 cm from the face
detection camera [7], [8]. Real-time facial recognition research has been conducted and has an accuracy result
of 87% using the convolutional neural network (CNN) method with 372 data points, including 186 male image
data and 186 female image data [9], [10]. Another research used 1,100 images, with 550 being male image data
and 550 female image data. Producing an accuracy value of 90% for age and 96% for gender using the CNN
method [11]. Research entitled "optimizing gender classification accuracy in facial images using data
augmentation and inception V-3" using the inception V-3 and convolutional methods with 290 data consisting
of 145 men and 145 women produced the best accuracy value of 99.31% [12], [13].
Although facial recognition technology has been advanced, most of the systems created are limited
to theoretical validation or rely on high-performance GPUs. The focus of this research is on the efficiency of
limited hardware and resources by developing a lightweight and accessible facial recognition system.
This system is specifically designed for home security. The distinguishing value of this research is the
implementation of digital image processing for a security system with a high level of accuracy. Digital image
processing can identify and recognize recorded faces precisely. By using this system, access control in a
room in the house can avoid the risk of hacking and access by faces that are not recognized by the system, so
that it can be secured [14]. The use of the digital image processing method has been proven to provide results
with a high level of accuracy, according to research that has been conducted on digital image processing,
which can recognize facial images better than other methods that have been tested [15]. Therefore, the use of
this method will be very suitable for implementation as a form of room security technology that requires
variations in facial expressions to uncertainty in changes in appearance [16].
This research was conducted to determine how the application of the digital image processing
method as facial recognition in home security systems and whether digital image processing can produce a
good level of accuracy for human facial recognition. It is expected that this technology can distinguish
between known and unknown faces so that only confirmed faces can be authorized to enter the area.
The system does not require special devices so that it is affordable and increases accessibility in its
application, especially in residential areas that require more security. The purpose of this research is to
implement a facial recognition system using digital image processing and deep learning tools based on
Python and OpenCV then evaluate the accuracy and robustness of the system in real-world constraints
including lighting variations and various facial poses. Another objective is to demonstrate the feasibility of
implementing the system on low-cost hardware for residential security use and system performance will be
measured through quantitative metrics including accuracy, precision, recall, F1 score, and confusion
matrix analysis.

2. METHOD
The face recognition system using digital image processing requires analysis to cover various
variations of human faces from various practical conditions. These conditions may occur in the use of the
system and must be addressed properly. Some of the analyses that need to be carried out by digital image
processing include data analysis, process analysis, system design, and system implementation.

2.1. Data analysis

The data source consists of a dataset of human facial images trained through image samples with
various poses, expressions, lighting, and different backgrounds. The purpose of this variation is for digital image
processing to be able to understand and recognize registered faces from various conditions. The total amount of
trained image data is adjusted to the number of subjects involved and to the desired variation. Furthermore, test
image data and performance tests of trained data are carried out. This test is carried out to evaluate the reliability
of the system in recognizing faces that have been registered by digital image processing [17].

2.2. Process analysis

This process performs an in-depth analysis related to the face recognition process in digital image
processing by considering the application built and the specifications of the registered dataset. The first
stage is image acquisition, where data that has been entered into the system can detect people's faces using
a real-time camera. Each face detected by the system is identified and named. The success of the system
can be indicated by the appearance of a name on the face that has been registered, and the camera can
consistently detect individual faces quickly and accurately from various conditions [18]. The next stage is
the face detection stage, face detection using digital image processing assisted by the OpenCV library [19],
[20]. This process is carried out to find the presence of a face in the camera frame. This face detection uses
the concept of Haar features, integral image, and cascade classifier so that the face detection process runs
efficiently [21].

Comput Sci Inf Technol, Vol. 7, No. 1, March 2026: 1-9

Comput Sci Inf Technol ISSN: 2722-3221  3

The third stage is pre-processing, which is the process of improving image quality before the next
process is continued. This process requires cropping on the target face only and removing irrelevant areas, then
the image resizing process so that when the image is converted to grayscale mode it is easy to extract and reduce
noise and contrast adjustments and when changing the color to grayscale are done to focus information for more
efficient and accurate face recognition. The next is the feature extraction stage, taking important features based
on the registered data to distinguish between faces. The extracted features include visual patterns that represent
unique features of the registered face [22]. The final stage is face matching by measuring the level of similarity
between the training image and the test image with a digital image processing method. This process compares
the characteristics learned by the system, so that it can accurately detect the visuals seen.

2.3. System design

The design of a facial recognition system requires several considerations, such as system
architecture design, a facial recognition model that meets specifications, a camera for facial image
acquisition, software for implementing facial recognition algorithms, and its interface. The system design has
a facial detection process starting with image input, followed by a pre-processing process. After that, the
feature extraction process, classification process, and image output are carried out. The application of face
recognition for home security systems begins by capturing facial images with the camera and automatically
converting them from RGB images to grayscale images. The use of the Haar cascade classifier method to
detect faces captured by the camera is marked by a blue circle on the face that has been detected [23].
The contents of the dataset are various facial images that represent subjects from various conditions
when taking pictures, so that the system can recognize faces as a whole, up to various situations that occur.
The design of this system requires data pre-processing such as cropping, resizing, and converting to
grayscale. This process needs to be done to improve the accuracy of facial recognition [24]. The flow of
dataset creation is as follows in Figure 1.
The face detected by the camera will be known by its histogram. The system will compare the
histogram value of the face detected by the camera with that in the database. The face will be recognized if its
histogram value is most similar to the data. By applying the local binary pattern histogram algorithm, the
camera can automatically compare histograms on the face [25]. The face recognition flow is shown in Figure 2.
The design of digital image processing applications for facial recognition systems has important steps for its
management, such as dataset separation, processing learning, prediction, evaluation, and digital image
processing architecture design [26], [27].

Figure 1. Dataset creation flow

Figure 2. Face recognition flow

Implementation of face recognition using Python (Febrian Wahyu Christanto)
4  ISSN: 2722-3221

2.4. System implementation

Dataset separation is the beginning of the implementation of a system consisting of several facial
images that are separated into two parts, namely training data and test data. Training data will be used to train
the digital image model, while test data will be used to test the performance of the trained model. Then the
digital image processing process is carried out using training data to train digital images. The process
involves forward propagation and backward propagation to optimize model parameters on facial patterns.
Digital images are trained to make predictions on newly seen facial images. The process is carried
out by classifying previously unseen facial images that have been learned from digital images. The results
will later be used to identify related individuals. Evaluation is the final stage to measure the performance of
the digital image model. Some of the evaluation measures are accuracy, precision, recall, and F1 score. This
evaluation is carried out to assess the level of sensitivity of digital images in recognizing faces accurately or
not. Haar Cascade excels in speed and simplicity but multi-task cascaded convolutional network (MTCNN)
or YOLO are superior for complex and high-precision applications especially in poor lighting conditions,
rotations, or varying expressions.

3. RESULTS AND DISCUSSION

The hardware implementation environment used has the following specifications, including
processor: Intel Celeron N4020 (2C/2T, 1.1/2.8GHz, 4MB), screen: 14″ HD (1366×768) TN 220nits
anti-glare, memory: 4GB DDR4-2400MHz, 256GB SSD M.2 2242 PCIe 3.0×2 NVMe storage, integrated
Intel UHD Graphics, and 2 cells 35WH35 battery.
Next is the software implementation environment used, such as the programming language
used is Python 3.6, which has flexibility and many supporting libraries in image processing. Python has an
easy-to-understand syntax and ease in utilizing various latest tools and techniques in application
development. The libraries and frameworks used in this study are TensorFlow: 2.16; Keras: 3.0.5; Scikit-
Learn: 1.4.1; Pandas: 2.2.1; OpenCV: [Link]; Matplotlib: 3.8.3; and NumPy: 1.26.4. development tools
using Jupyter Notebook, Spyder, and Tkinter. The right combination of programming languages and libraries
will make it easy to develop applications with satisfactory and efficient results.
Implementation of facial recognition is done by separating the dataset according to the name that
has been registered in the system. Grouping according to this data is done based on the face of each
individual, who will create a folder containing sample images of their face. The create_directory function
will create a folder that will divide the type of face. Furthermore, the system will perform an OpenCV face
detector using a webcam. The detected image will be stored in the directory according to the name registered
to distinguish the sample. After the detected face enters the directory, the next step is processing the dataset
for faces that have been categorized by name. The images that have been entered are converted to grayscale
to reduce the data dimension, but still get the required data, namely the visual characteristics of the facial
image. Before being converted to grayscale, the image will be cropped so that only the face part is obtained as
the required data, then the image that has gone through the cropping process will be resized to 100×100 pixels
so that the size is uniform and by the digital image model. Only after that are the images converted to
grayscale to facilitate analysis and efficiency of the digital image model, so that the dataset is ready to be
used in facial recognition, as in Figure 3.
Next, the face recognition system using digital images will recognize known faces according to
those registered in the OpenCV library. The recognition system can directly use the camera in real-time.
The results of this recognition will show the name of the registered face user continuously, as in Figure 4.
The creation of the basic concept of digital image architecture, covering the main layers in the
formation of digital image structures such as convolutional layers, pooling layers, and fully connected layers,
is carried out at this stage. Each layer will extract from each facial recognition feature and combine them into
one unit for accurate prediction results. The first layer is a convolution (Conv2D) with 32 filters measuring
3×3 with the rectified linear unit (ReLU) activation function. Image input uses a color image with dimensions
of 100×100 pixels. Followed by a MaxPooling layer (MaxPooling2D) measuring 2×2 to reduce the image
dimension. After both of these things run, the results will be done for flattening (dlatten), which changes the
vector to dimensions. The vector will be forwarded to the dense layer with 128 units and uses the ReLU
activation function. The dense report provides output according to the number of faces recognized using the
softmax activation function to produce prediction probabilities for each class.
The description and rationale for parameter selection are from 32 filters and 3×3 kernel with small
size and efficient for initial feature detection because 32 filters are sufficient for small datasets. MaxPooling
2×2 commonly used to reduce the size of the feature map while preserving important features. In addition,
ReLU serves as a popular activation function because it avoids the vanishing gradient problem and softmax
used for multiclass classification in the output layer.

Comput Sci Inf Technol, Vol. 7, No. 1, March 2026: 1-9

Comput Sci Inf Technol ISSN: 2722-3221  5

The model training process uses the Adam optimization algorithm to produce optimal prediction
probabilities. The way the Adam method works is to update the parameters of the model based on the gradient
and loss function, so that periodically, the performance of the system's decision selection from its predictions
becomes better [28]. The results of the analysis show that the evaluation matrix obtained is accurate to help
determine how well the model matches facial images in each category. Then, the precision evaluation is
carried out to understand the reliability of the model in identifying facial images from its class [29]. Recall or
sensitivity level to measure the proportion of the predicted results to the actual positive data. F1-score is a
metric that combines precision and recall into one number. F1-score provides an overview of the quality of the
mode prediction and considers its balance. Evaluation results will provide insights related to the performance
of the digital image model. The data shown are accuracy, evaluation matrix, and heatmap confusion matrix.

Figure 3. Image dataset Figure 4. Facial recognition process

3.1. Training and validation accuracy graph

Accuracy and validation training process looks at how well the model learns the data and
generalizes to the test data. Changes in model accuracy are made along with the addition of epochs indicated
by the training accuracy on the blue line to show how well the model learns the training data will remembers
the patterns in the data. The orange line which is the validation accuracy shows how well the model
generalizes new data that is not in the training directory to avoid adjusting new data. The results of the
validation training and validation accuracy indicated by the blue and orange lines are depicted in a chart in
Figure 5. Based on the epoch versus accuracy graph, this model achieves a level of accuracy stability
reaching 3 epochs with a training accuracy reaching 95% with high and stable validation accuracy.

3.2. Evaluation matrix graphic

Scope of the evaluation matrix is accuracy, precision, recall, and F1-score. This evaluation helps to
find out the strengths and weaknesses of the model for clustering, as in Figure 6. Assessment of the
evaluation matrix shows a range of 0 to 1. If it shows the number 1, then the performance of the model is
optimal. The graph in Figure 5 shows the maximum value, which is 97.5%, with good performance in
recognizing and classifying facial images with a high level in each evaluation matrix.

Figure 5. Validation training accuracy chart

Implementation of face recognition using Python (Febrian Wahyu Christanto)

6  ISSN: 2722-3221

Figure 6. Evaluation matrix graphic

3.3. Heatmap confusion matrix

Evaluation table of the classification model performance in the row table is the actual label, and the
column table is the predicted label. The distribution of model predictions in color shows how well the model
predicts, or there are errors. There are four classes registered, namely Fadia, Vicky, Arifin, and Aisyah. The
matrix shows the value of the number of predictions formed by the model in each class and compared to the
actual class. The actual label class has 4 names registered in the model. Diagonally from left to right, the
number of model prediction values in predicting the correct label. In the Fadia class, it shows 6 correct
predictions, the Vicky class has 15 correct predictions, the Arifin class has 9 correct predictions, and the
Aisyah class has 10 correct predictions. In other cells that show incorrect predictions, there is only 1
prediction error. The color and scale in the table show that the dark color is the high frequency, and the scale
on the right shows the range of values associated with the matrix color. It can be concluded that this model
can predict accurately and effectively without any lack of variation in the test data as shown in Figure 7.

Figure 7. Heatmap confusion matrix

The testing process uses a graphical user interface (GUI) display program with an HTML-based
command script and is compared with the Python language to run the camera and digital image model
algorithm. The initial step is to enter the username to retrieve facial image data 100 times and then save it in a
folder according to the username. The next step is to test the registered face detection system, each of which
is tested 5 times. Based on the testing of the four user faces, each of which was tested five times and

Comput Sci Inf Technol, Vol. 7, No. 1, March 2026: 1-9

Comput Sci Inf Technol ISSN: 2722-3221  7

accumulated over 20 trials, only one detection error was found, namely in user Vicky in her third test as an
unknown face, as in Table 1. Cause of the failure of identification of the identification of the Vicky user is
due to lighting that is too bright or too dark, so that facial detection to identify moving faces has difficulty.
The level of accuracy of the results of this test is calculated using a percentage that produces 95% based on
19 out of 20 successful tests.

Table 1. Table results

Num Username Testing Identification results Status
1 Fadia 5 times Fadia 5 Successfully
2 Vicky 5 times 4 Vicky 4 Successfully
1 Unknown 1 Failed
3 Arifin 5 times Arifin 5 Successfully
4 Aisyah 5 times Aisyah 5 Successfully

4. CONCLUSION
The conclusion that can be drawn is that the accuracy of the digital image model reaches around
3 epochs. The training accuracy reaches 95% with high and stable validation accuracy. So that the model can
learn patterns in the data well and can generalize to the test data. The maximum value obtained is 97.5%
therefore, the model has succeeded in doing its job in recognizing and classifying faces well by showing high
levels of accuracy, precision, recall, and F1 scores. System analysis related to face recognition with the
algorithm method conducted 20 trials with 4 face names, as many as 400 images of different faces, getting a
program accuracy level of 95%, and from the overall recognition, there was only 1 error in recognition.
Performance needed for further development is better and more complex, which requires the need to use a
higher quality GPU and a higher RAM size so that the dataset can be expanded and the process of
understanding variations becomes better. The facial recognition system can also be tested again with certain
cases such as facial rotation, facial expression variations, different lighting, and age. Edge deployment
optimizes the model to run efficiently on Raspberry Pi, Jetson Nano, and others. Emotion detection can also
be added by adding an emotion classification module to detect suspicious expressions. Multi-face tracking
with ID for crowd recognition scenarios along with spoofing defense for liveness verification, blink
detection, or live skin texture can enhance this system.

ACKNOWLEDGMENTS
The research team extends its heartfelt gratitude to the Universitas Semarang, Satya Wacana
Christian University, Deakin University, and Bina Nusantara University for their invaluable support and
contribution to this research.

FUNDING INFORMATION
No funding is involved.

AUTHOR CONTRIBUTIONS STATEMENT

This journal uses the Contributor Roles Taxonomy (CRediT) to recognize individual author
contributions, reduce authorship disputes, and facilitate collaboration.

Name of Author C M So Va Fo I R D O E Vi Su P Fu
Febrian Wahyu ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Christanto
Husnul Arifin ✓ ✓ ✓ ✓ ✓ ✓
Christine Dewi ✓ ✓ ✓ ✓ ✓ ✓
Teguh Prasandy ✓ ✓ ✓ ✓ ✓ ✓

C : Conceptualization I : Investigation Vi : Visualization

M : Methodology R : Resources Su : Supervision
So : Software D : Data Curation P : Project administration
Va : Validation O : Writing - Original Draft Fu : Funding acquisition
Fo : Formal analysis E : Writing - Review & Editing

Implementation of face recognition using Python (Febrian Wahyu Christanto)

8  ISSN: 2722-3221

CONFLICT OF INTEREST STATEMENT

Authors state no conflict of interest.

DATA AVAILABILITY
No data are available for sharing as this research did not use an external dataset.

REFERENCES
[1] Z. S. Shouran, N. Rokhman, and T. K. Priyambodo, “System security awareness planning model using the octave method
approach,” IJCCS (Indonesian Journal of Computing and Cybernetics Systems), vol. 13, no. 3, 2019, doi: 10.22146/ijccs.43922.
[2] H. Fitriyah and E. R. Widasari, “Face detection of thermal images in various standing body-pose using facial geometry,” IJCCS
(Indonesian Journal of Computing and Cybernetics Systems), vol. 14, no. 4, Oct. 2020, doi: 10.22146/ijccs.59672.
[3] A. W. Dorn, S. Webb, and S. Pâquet, “From wargaming to peacegaming: digital simulations with peacekeeper roles needed,”
International Peacekeeping, vol. 27, no. 2, pp. 289–310, Mar. 2020, doi: 10.1080/13533312.2020.1721287.
[4] A. Vatresia, R. Faurina, V. Purnamasari, and I. Agustian, “Automatic fish identification using single shot detector,” CommIT
Journal, vol. 16, no. 2, pp. 167–174, Jun. 2022, doi: 10.21512/commit.v16i2.8126.
[5] J. Ni, R. Wang, and J. Tang, “ADSSD: improved single-shot detector with attention mechanism and dilated convolution,” Applied
Sciences, vol. 13, no. 6, Mar. 2023, doi: 10.3390/app13064038.
[6] S. Sunardi, A. Fadlil, and D. Prayogi, “Face recognition using machine learning algorithm based on Raspberry Pi 4b,”
International Journal of Artificial Intelligence Research, vol. ISSN, no. 1, pp. 2579–7298, 2022.
[7] G. Wang, J. Li, Z. Wu, J. Xu, J. Shen, and W. Yang, “EfficientFace: an efficient deep network with feature enhancement for
accurate face detection,” Multimedia Systems, vol. 29, no. 5, pp. 2825–2839, Oct. 2023, doi: 10.1007/s00530-023-01134-6.
[8] Aballe et al., “Security measures: effectiveness of the installation of CCTV cameras in relation to crime prevention as perceived
by the community,” Middle East Journal of Applied Science & Technology, vol. 05, no. 02, pp. 149–160, 2022,
doi: 10.46431/mejast.2022.5216.
[9] I. H. Al Amin, D. E. Marinda, E. Winarno, U. N. D. Handayani, and V. Lusiana, “Real-time detection of face mask using
convolutional neural network,” Jurnal RESTI, vol. 7, no. 3, pp. 697–704, Jun. 2023, doi: 10.29207/resti.v7i3.5036.
[10] M. K. Benkaddour, S. Lahlali, and M. Trabelsi, “Human age and gender classification using convolutional neural network,” in
2020 2nd International Workshop on Human-Centric Smart Environments for Health and Well-Being, IHSH 2020, Feb. 2021,
pp. 215–220. doi: 10.1109/IHSH51661.2021.9378708.
[11] N. Kashyap, A. Mann, and N. Sharma, “Age and gender detection system,” SSRN Electronic Journal, 2024,
doi: 10.2139/ssrn.4931530.
[12] J. P. Tanjung, M. R. Faldi, H. Sitompul, M. Ridho, and J. P. Ambarita, “Optimizing gender classification accuracy in facial
images using data augmentation and inception V-3,” Sinkron, vol. 8, no. 4, pp. 2627–2634, Oct. 2023,
doi: 10.33395/sinkron.v8i4.12785.
[13] S. M. Daniyal, A. Masood, M. Ebrahim, M. Ebrahim, S. H. Adil, and K. Raza, “An improved face recognition method based on
convolutional neural network,” Journal of Independent Studies and Research Computing, vol. 22, no. 1, 2024,
doi: 10.31645/JISRC.[Link].
[14] R. Matulionyte and M. Zalnieriute, The Cambridge handbook of facial recognition in the modern state. Cambridge University
Press, 2024. doi: 10.1017/9781009321211.
[15] K. N. Singh, O. P. Singh, and A. K. Singh, “ECiS: Encryption prior to compression for digital image security with reduced
memory,” Computer Communications, vol. 193, pp. 410–417, Sep. 2022, doi: 10.1016/[Link].2022.07.049.
[16] H. L. Gururaj, B. C. Soundarya, S. Priya, J. Shreyas, and F. Flammini, “A comprehensive review of face recognition techniques ,
trends, and challenges,” IEEE Access, vol. 12, pp. 107903–107926, 2024, doi: 10.1109/ACCESS.2024.3424933.
[17] W. Xue, Z. Zheng, F. Lu, H. Wei, G. Chen, and C. Jiang, “GeoNLF: geometry guided pose-free neural LiDAR fields,” Advances
in Neural Information Processing Systems, vol. 37, 2024.
[18] A. B. Shetty, Bhoomika, Deeksha, J. Rebeiro, and Ramyashree, “Facial recognition using Haar cascade and LBP classifiers,”
Global Transitions Proceedings, vol. 2, no. 2, pp. 330–335, Nov. 2021, doi: 10.1016/[Link].2021.08.044.
[19] A. Willis and K. Hasan, “Simple computer vision algorithm production using OpenCV for "virtual ecosystem” project,” IJNMT
(International Journal of New Media Technology), vol. 7, no. 1, pp. 23–27, Jul. 2020, doi: 10.31937/ijnmt.v7i1.1660.
[20] K. Burugu, V. Dodda, V. Rajmoor, S. C. Yerra, and S. Bhutada, “Security surveillance monitoring system using Open CV,”
International Journal of Science Research in Computer Science, Engineering and Information Technology, vol. 8, no. 3,
pp. 402–407, 2022.
[21] R. B. Y. Al-Khafaji, “Using technology and algorithms for face detection and recognition using digital image processing and
relying on a computer vision sensor,” Journal of Information Systems Engineering and Management, vol. 9, no. 1, Jan. 2024,
doi: 10.55267/iadt.07.14328.
[22] H. Putri et al., “Security system for door locks using YOLO-based face recognition,” International Journal on Informatics
Visualization, vol. 9, no. 1, pp. 224–230, Jan. 2025, doi: 10.62527/joiv.9.1.2410.
[23] H. J. Mun and M. H. Lee, “Design for visitor authentication based on face recognition technology using CCTV,” IEEE Access,
vol. 10, pp. 124604–124618, 2022, doi: 10.1109/ACCESS.2022.3223374.
[24] B. Koodalsamy, M. B. Veerayan, and V. Narayanasamy, “Face recognition using deep learning,” E3S Web of Conferences,
vol. 387, May 2023, doi: 10.1051/e3sconf/202338705001.
[25] M. H. Abdurrahman, H. Amran Darwito, and A. Saleh, “Face recognition system for prevention of car theft with Haar cascade
and local binary pattern histogram using Raspberry Pi,” EMITTER International Journal of Engineering Technology,
pp. 407–425, Dec. 2020, doi: 10.24003/emitter.v8i2.534.
[26] X. Fu, “Design of facial recognition system based on visual communication effect,” Computational Intelligence and
Neuroscience, vol. 2021, no. 1, Jan. 2021, doi: 10.1155/2021/1539596.
[27] X. Liu and R. Yao, “Design of visual communication teaching system based on artificial intelligence and CAD technology,”
Computer-Aided Design and Applications, vol. 20, no. S10, pp. 90–101, Apr. 2023, doi: 10.14733/cadaps.2023.S10.90-101.

Comput Sci Inf Technol, Vol. 7, No. 1, March 2026: 1-9

Comput Sci Inf Technol ISSN: 2722-3221  9

[28] Y. Shao et al., “An improved BGE-Adam optimization algorithm based on entropy weighting and adaptive gradient strategy,”
Symmetry, vol. 16, no. 5, May 2024, doi: 10.3390/sym16050623.
[29] N. Xiao, X. Hu, X. Liu, and K.-C. Toh, “Adam-family methods for nonsmooth optimization with convergence guarantees,”
Journal of Machine Learning Research, vol. 25, pp. 1-53, 2024.

BIOGRAPHIES OF AUTHORS

Febrian Wahyu Christanto received the B.S. degree in Informatic Engineering

in 2010 and his M.S. in Information System in 2012 respectively from the Satya Wacana
Christian University, Salatiga, Indonesia. Since 2013 he has been an Assistant Professor at the
Information Technology Department of Universitas Semarang, Indonesia. He has published
several journal and conference articles in his areas and active as a reviewer and guest editor in
many journals. His research interests include computer networking, internet of things, and
software engineering. He can be contacted at email: [Link]@[Link].

Husnul Arifin received the Bachelor of Informatics Engineering ([Link].) from

the Department of Information Technology Universitas Semarang, Indonesia in 2024. His
research topics include artificial intelligence and object detection and recognition. He can be
contacted at email: arifinhusnul20@[Link].

Christine Dewi received a B.S. Degree from the Informatics Engineering study
program in 2010, and a Master of Computer Science from the Master of Information Systems
study program in 2012, both from the Satya Wacana Christian University, Salatiga, Indonesia.
In 2021 she finished her Ph.D. and work as a postdoctoral researcher at Chaoyang University
of Technology, Taiwan until 2022. She is now an Associate Professor in the Department of
Information Technology of Satya Wacana Christian University, Indonesia and Researcher at
Deakin University, Australia. She is active as a reviewer and guest editor in many journals.
Her current research interests include image processing, computer vision, object detection and
recognition, artificial intelligence, and machine learning. She can be contacted at email:
[Link]@[Link] or [Link]@[Link].

Teguh Prasandy working as an Online Lecturer Coordinator in the Business

Competitive Intelligence Group at the Information System Department, Bina Nusantara
University, Indonesia. Graduated from Water Resources Management, Diponegoro University
in 2007. Graduated from the Master of Information Systems, Diponegoro University in 2011.
and graduated from Informatics Engineering, Universitas Semarang in 2016. Interested in
research on web-based applications for management decision making. He can be contacted at
email: [Link]@[Link].

Implementation of face recognition using Python (Febrian Wahyu Christanto)