Unraveling Indonesian heritage through pattern recognition using YOLOv5

Computer Science and Information Technologies
Vol. 5, No. 3, November 2024, pp. 265~271
ISSN: 2722-3221, DOI: 10.11591/csit.v5i3.pp265-271  265
Journal homepage: http://iaesprime.com/index.php/csit
Unraveling Indonesian heritage through pattern recognition
using YOLOv5
Rosalina1
, Genta Sahuri2
1
Informatics Study Program, Faculty of Computing, President University, Indonesia
2
Information Systems Study Program, Faculty of Computing, President University, Indonesia
Article Info ABSTRACT
Article history:
Received Aug 9, 2024
Revised Sep 18, 2024
Accepted Oct 2, 2024
This research focuses on three iconic Indonesian batik patterns-Kawung,
Mega Mendung, and Parang-due to their cultural significance and
recognition. Kawung symbolizes harmony, Mega Mendung represents
power, and Parang signifies protection and spiritual power. Using the
YOLOv5 deep learning model, the study aimed to accurately identify these
patterns. Results showed mean average precision (mAP) scores of 77% for
Kawung, 80% for Parang, and an impressive 99% for Mega Mendung. The
highest precision results were 91% for Kawung, 88% for Parang, and 77%
for Mega Mendung. These findings highlight the potential of pattern
recognition in preserving cultural heritage. Understanding these designs
contributes to the appreciation of Indonesia s culture. The research suggests
applications in cultural studies, digital archiving, and the textile industry,
ensuring the legacy of these patterns endures.
Keywords:
Cultural heritage
Indonesian batik
Pattern recognition
Preservation
YOLOv5
This is an open access article under the CC BY-SA license.
Corresponding Author:
Rosalina
Informatics Study Program, Faculty of Computing, President University
Jl. Ki Hajar Dewantara, Cikarang - Bekasi, Indonesia
Email: rosalina@president.ac.id
1. INTRODUCTION
Batik, an ancient art form, holds immense cultural significance in Indonesia, reflecting the country s
rich heritage and diverse traditions. Among the myriad of batik patterns, Kawung, Mega Mendung, and
Parang stand out as iconic motifs deeply rooted in Indonesian culture. These patterns are not merely
decorative; they carry profound meanings and stories that connect generations. However, as the world
transitions into the digital age, there is a pressing need to preserve and promote these cultural treasures using
modern technologies. Batik, a UNESCO Intangible Cultural Heritage, is more than just a fabric; it is a
symbol of Indonesian identity and craftsmanship [1]. Each batik pattern is a reflection of Indonesia s diverse
cultural heritage, with designs often inspired by nature, mythology, and spiritual beliefs. Kawung, for
example, is believed to be one of the oldest batik motifs, symbolizing harmony and balance [2]. Mega
Mendung, originating from Cirebon, features cloud motifs and represents power and strength [3]. Parang,
with its dagger-like motifs, is believed to provide protection and spiritual power.
In recent years, there has been a growing interest in digitizing batik patterns to preserve them for
future generations [4]-[17]. However, digitizing these intricate designs poses several challenges, including
the need for accurate pattern recognition [18]–[20]. Manual recognition methods are time-consuming and
prone to errors, highlighting the necessity for automated pattern recognition systems [21]–[28]. This research
proposes a novel approach to address this gap by utilizing the YOLOv5 deep learning model for pattern
recognition. YOLOv5 is renowned for its speed and accuracy in object detection tasks, making it a promising
candidate for recognizing complex batik patterns. By training the model on a dataset comprising images of

 ISSN: 2722-3221
Comput Sci Inf Technol, Vol. 5, No. 3, November 2024: 265-271
266
Kawung, Mega Mendung, and Parang, we aim to achieve high accuracy in identifying these patterns,
providing a valuable tool for researchers, artisans, and enthusiasts alike.
One of the challenges in preserving batik heritage lies in the accurate identification and
differentiation of these intricate patterns, especially as they can vary in subtle ways across regions and
artisans. Manual recognition methods are time-consuming and prone to errors, highlighting the necessity for
automated pattern recognition systems. While some efforts have been made to digitize batik patterns, there
remains a lack of comprehensive studies focusing on the recognition of specific iconic designs like Kawung,
Mega Mendung, and Parang. This research proposes a novel approach to address this gap by utilizing the
YOLOv5 deep learning model for pattern recognition. YOLOv5 is renowned for its speed and accuracy in
object detection tasks, making it a promising candidate for recognizing complex batik patterns. By training
the model on a dataset comprising images of Kawung, Mega Mendung, and Parang, we aim to achieve high
accuracy in identifying these patterns, providing a valuable tool for researchers, artisans, and enthusiasts
alike. This research is novel in its focus on culturally significant batik patterns and its use of a state-of-the-art
deep learning model for recognition. While previous studies have primarily explored general batik
digitization, this research takes a more nuanced approach by targeting patterns with special meaning for
Indonesians. The proposed solution not only contributes to preserving Indonesia s cultural heritage but also
demonstrates AI s potential in cultural studies and digital archiving, setting a precedent for future research in
the field.
2. METHOD
In this study, we classify three types of Indonesian batik motifs using the YOLOv5 PyTorch format.
The process involves dataset collection, data preprocessing, model training, and inference (Visually
represented in Figure 1). Users input images from photos, videos, or live webcam feeds, which are then
matched with pre-trained data. The model is trained to accurately classify batik motifs as Kawung, Mega
Mendung, or Parang. Evaluation metrics like mean average precision (mAP) are used to assess the model s
performance. This methodology ensures a reliable classification system for Indonesian batik motifs,
preserving their cultural value in the digital era.
Figure 1. Steps in classifying three types of Indonesian Batik motifs
2.1. Dataset collection
The process of preparing the dataset for classifying Indonesian batik motifs involved several steps to
ensure its quality and suitability for training the model. Initially, we sourced images containing the three
specific batik motifs (Kawung, Parang, and Mega Mendung) from publicly available photos on Google.
These images were then compiled and organized using Roboflow, a platform designed for managing
computer vision datasets.
In this research, the image dataset is categorized into three groups. There are 143 annotated photos
of the Kawung batik motif. Additionally, there are 101 annotated photos each for the Parang and Mega
Mendung batik motifs. In total, we have collected 345 annotated photos. These images are sourced from
Google and have been annotated for classification purposes (llustrated in Figure 2).
2.2. Preprocessing and augmentation
Once the images were collected, they underwent annotation, labeling, and resizing using Roboflow s
tools. Annotation involved marking the specific areas within each image that contained the batik motifs,
ensuring that the model would learn to recognize these patterns accurately. Labeling helped categorize the
images into their respective classes (kawung, parang, and mega mendung), which is crucial for training a

Comput Sci Inf Technol ISSN: 2722-3221 
Unraveling Indonesian heritage through pattern recognition using YOLOv5 (Rosalina)
267
classification model. After annotation and labeling, the dataset was split into three sets: training, validation,
and test sets. The training set, comprising approximately 89% of the images, was used to train the model. The
validation set, accounting for 6% of the images, was used to tune the model s hyperparameters and evaluate
its performance during training. The test set, consisting of 4% of the images, was used to assess the final
performance of the trained model.
To improve the model s ability to generalize and recognize batik motifs in various conditions, we
performed data augmentation. This involved applying a series of transformations to the images, such as
flipping (both horizontally and vertically), rotating (both clockwise and counter-clockwise by 90°), and
adjusting brightness. These transformations help the model learn to recognize batik motifs from different
angles, orientations, and lighting conditions, making it more robust and capable of accurately classifying
batik patterns in real-world scenarios.
Figure 2. Sample of dataset
2.3. Convert data to YOLOV5
YOLOv5, short for "You Only Look Once, Version 5," is an advanced algorithm used for efficient
and real-time object detection. It is specifically designed to quickly identify objects based on the categories
they have been trained on. This algorithm is known for its speed and accuracy, making it a popular choice for
various computer vision applications. One of the key advantages of YOLOv5 is its open-source nature, which
allows users to easily access and modify the code according to their specific needs. This means that users can
create custom models tailored to detect specific objects, making it a versatile tool for object detection tasks.
Converting datasets into YOLOv5 format involves several steps to ensure compatibility with the
algorithm. First, the images in the dataset need to be resized to a suitable input size for the YOLOv5 model.
Typically, YOLOv5 requires images to be resized to a square shape, such as 640×640 pixels. Next, the
annotations for each image, which contain information about the location and class of objects in the image,
need to be converted into a specific format that YOLOv5 understands. This format typically includes the
class label, followed by the bounding box coordinates normalized to the range [0, 1], relative to the width and
height of the image. Once the images and annotations are prepared, they can be converted into the YOLOv5
format using a script or tool provided by the YOLOv5 framework. This process involves creating text files
for each image, where each line contains the class label and bounding box coordinates for each object in the
image. These text files are then used as input to train the YOLOv5 model.
2.4. Training and testing
In the model configuration stage, we tailor the YOLOv5 model for object detection to suit our
specific task of recognizing Indonesian batik motifs. One crucial parameter to set is the number of classes,
which in this case is 3 to represent the three batik motifs: Kawung, Mega Mendung, and Parang. This tells the
model how many different types of objects it should be looking for in the images. Additionally, we set the
learning rate, which controls how quickly the model adapts its weights during training. A higher learning rate
can lead to faster convergence but may also cause the model to overshoot the optimal weights. Conversely, a
lower learning rate may lead to slower convergence but may result in more stable training. The batch size,
which determines the number of images processed in each iteration of training, is also specified. A larger
batch size can lead to faster training but may require more memory.

 ISSN: 2722-3221
268
In the model training stage, we use the configured YOLOv5 model to train on the prepared training
set. We employ an optimizer like Adam, which adapts the learning rate for each parameter during training to
improve convergence. The loss function used combines localization loss and classification loss. Localization
loss measures the accuracy of the predicted bounding boxes compared to the ground truth boxes.
Classification loss measures how accurately the model predicts the class of each object. By combining these
two components, the model learns to both localize and classify objects in the images. Throughout training,
we monitor the model s performance on the validation set. This helps us to detect signs of overfitting, where
the model learns to memorize the training data instead of generalizing to new, unseen data. If overfitting is
detected, we can adjust hyperparameters or introduce regularization techniques to improve the model s
performance.
2.5. Validation
In the validation stage, the trained YOLOv5 model is evaluated using the validation set, which
consists of images that were not seen during training. This evaluation is crucial for assessing how well the
model generalizes to new, unseen data and for fine-tuning hyperparameters to improve performance. First,
the validation set images are preprocessed, similar to the training set, by resizing them to a suitable input size
for the YOLOv5 model and applying any necessary augmentation techniques. Next, the model is used to
make predictions on the validation set images. It processes each image and produces bounding boxes and
class predictions for the objects it detects. To quantify the model s performance, metrics such as precision,
recall, and mAP are calculated. Precision measures the proportion of correctly predicted positive instances
out of all instances predicted as positive. Recall measures the proportion of true positive instances that are
correctly predicted out of all actual positive instances. mAP considers the average precision for each class,
providing an overall measure of detection performance across all classes. Based on the model s performance
on the validation set, hyperparameters such as the learning rate, batch size, and augmentation strategies may
be adjusted to improve performance. This iterative process of training, evaluation, and hyperparameter tuning
continues until satisfactory performance is achieved. The validation stage helps ensure that the YOLOv5
model is robust and performs well on unseen data, setting the stage for final testing and deployment.
3. RESULTS AND DISCUSSION
Results and discussions are crucial in evaluating the performance of the YOLOv5 model for
detecting and classifying Indonesian batik motifs. These results provide insights into the model s
effectiveness and its potential impact on preserving and promoting Indonesian cultural heritage. The results
of the model evaluation on the test set indicate its ability to accurately detect and classify the batik motifs.
The model achieved a mAP of 85% for Kawung batik, a result attributed to the motif s distinct and repetitive
nature, aiding the model s effective pattern recognition. Its consistent performance across various images
indicates its ability to handle design variations, establishing it as a reliable tool for identifying this traditional
motif (depicted in Figure 3).
Figure 3. The Precision, Recall, and mAP results for the Batik Kawung batik motif
In comparison, Parang batik achieved a mAP of 80%, slightly lower but still respectable. The
complexity of the Parang motif, with its detailed elements, may pose detection challenges, yet the 80% mAP
demonstrates the model s accurate recognition capability. Mega Mendung batik achieved an impressive 99%

269
mAP, indicating the model s effective learning of this bold and distinct design (depicted in Figure 4). The
motif s clear and repetitive nature likely contributes to the model s high precision in recognizing it. This
achievement showcases the potential of advanced machine learning in preserving and promoting iconic
cultural patterns.
Figure 4. The precision, recall, and mAP results for the Mega Mendung Batik motif
The model s performance can be further illustrated with examples. For instance, when presented
with an image containing the Kawung motif, the model successfully detects the motif and classifies it as
Kawung with high accuracy. Similarly, when presented with images containing the Mega Mendung and
Parang motifs, the model demonstrates its ability to accurately identify and classify these motifs. These
results are promising as they indicate that the YOLOv5 model is effective in recognizing Indonesian batik
motifs. This has significant implications for preserving and promoting Indonesian cultural heritage, as the
model can be used to identify and catalog batik motifs in digital archives, museums, and cultural institutions.
The discussions surrounding these results focus on the implications and limitations of the model.
While the results demonstrate the model s effectiveness, there are several limitations to consider. One
limitation is the size and diversity of the dataset. A larger and more diverse dataset would likely improve the
model s performance and generalization to unseen data. Another limitation is the complexity of the batik
motifs themselves. Some motifs may be more challenging to detect and classify due to their intricate designs
and variations. Additionally, the model s performance may be affected by factors such as lighting conditions,
image quality, and background clutter. Despite these limitations, the results of this research are promising
and indicate the potential of using deep learning models like YOLOv5 for recognizing and preserving
cultural heritage. Future research could focus on expanding the dataset, refining the model architecture, and
exploring other deep learning techniques to further improve the model s performance.
4. CONCLUSION
This research demonstrates the effectiveness of the YOLOv5 model in detecting and classifying
three iconic Indonesian batik motifs: Kawung, Parang, and Mega Mendung. The model achieved impressive
mAP scores, with 85% for Kawung, 80% for Parang, and a remarkable 99% for Mega Mendung, reflecting
its capability to accurately recognize these intricate patterns. The study underscores the importance of
leveraging modern machine learning techniques to preserve and promote cultural heritage. By digitizing and
automating the identification of traditional motifs, this approach offers a valuable tool for cultural
preservation, educational purposes, and global promotion of Indonesian batik. The model s robustness and
high accuracy highlight its potential for real-world applications, such as digital archiving and interactive
educational tools. Future research could enhance these results by expanding the dataset, incorporating more
diverse images, and experimenting with different augmentation techniques. Integrating the model into mobile
or web applications could further extend its reach, providing real-time pattern recognition and fostering
greater appreciation for Indonesia s rich cultural heritage.
ACKNOWLEDGEMENTS
The author gratefully acknowledges the support received from the Research and Community Service
Institute (LRPM) at President University, through the research grant program. This financial support was
crucial for the successful completion of this study.

 ISSN: 2722-3221
270
REFERENCES
[1] P. Priyanto, “Utilization of batik cultural arts for human resources development,” Journal of Indonesian Tourism and Policy
Studies, vol. 8, no. 2, p. 1, 2024, [Online]. Available: https://scholarhub.ui.ac.id/jitps/vol8/iss2/1.
[2] P. Nimas Ayu, N. A. Pramesti, H. Pamadhi, and A. Garbo, “Local wisdom values in Kawung Batik and its relevance to moral
education,” 2019, doi: 10.2991/icaae-18.2019.34.
[3] D. Trihanondo, D. Endriawan, T. Haryotedjo, G. M. Putra, and R. Machfiroh, “Redefining Cirebon Batik into an environmentally
friendly icon of West Java,” IOP Conference Series: Materials Science and Engineering, vol. 1098, no. 5, p. 052011, 2021, doi:
10.1088/1757-899x/1098/5/052011.
[4] R. Andriansyah et al., “Augmented reality approach to introduce Batik Garutan using features accelerated segment test,”
Proceeding of 2021 7th International Conference on Wireless and Telematics, ICWT 2021, 2021, doi:
10.1109/ICWT52862.2021.9678417.
[5] E. Winarno, W. Hadikurniawati, A. Septiarini, and H. Hamdani, “Analysis of color features performance using support vector
machine with multi-kernel for batik classification,” International Journal of Advances in Intelligent Informatics, vol. 8, no. 2, pp.
151–164, 2022, doi: 10.26555/ijain.v8i2.821.
[6] B. S. Negara, E. Satria, S. Sanjaya, and D. R. Dwi Santoso, “ResNet-50 for Classifying Indonesian Batik with Data
Augmentation,” 2021 International Congress of Advanced Technology and Engineering, ICOTEN 2021, 2021, doi:
10.1109/ICOTEN52080.2021.9493488.
[7] A. E. Minarno, Y. Azhar, F. D. Setiawan Sumadi, and Y. Munarko, “A robust batik image classification using multi texton co-
occurrence descriptor and support vector machine,” 2020 3rd International Conference on Intelligent Autonomous Systems,
ICoIAS 2020, pp. 51–55, 2020, doi: 10.1109/ICoIAS49312.2020.9081833.
[8] A. H. Rangkuti, A. Harjoko, and A. Putra, “A novel reliable approach for image batik classification that invariant with scale and
rotation using MU2ECS-LBP algorithm,” Procedia Computer Science, vol. 179, pp. 863–870, 2021, doi:
10.1016/j.procs.2021.01.075.
[9] A. D. Wibawa, E. Arif Wicaksono, S. D. Suryani, and R. Rumadi, “Javanese Batik image classification using self-organizing
map,” ICCoSITE 2023 - International Conference on Computer Science, Information Technology and Engineering: Digital
Transformation Strategy in Facing the VUCA and TUNA Era, pp. 472–477, 2023, doi: 10.1109/ICCoSITE57641.2023.10127783.
[10] A. F. Ihsan, “Initial study of batik generation using variational autoencoder,” Procedia Computer Science, vol. 227, pp. 785–794,
2023, doi: 10.1016/j.procs.2023.10.584.
[11] R. Mawan, K. Kusrini, and H. Al Fatta, “The effect of image dimensions on batik motif classification using convolutional neural
networks" (in Indonesian) "Pengaruh dimensi gambar pada klasifikasi motif batik menggunakan convolutional neural network,”
Jurnal Teknologi Informasi, vol. 4, no. 2, pp. 218–223, 2020, doi: 10.36294/jurti.v4i2.1342.
[12] I. M. Ihdal, “Classification of typical batik fabrics and typical sasirangan fabrics using the convolutional neural network method"
(in Indonesian) "Klasifikasi kain khas batik dan kain khas sasirangan dengan menggunakan metode convolutional neural
network,” Jurnal Teknologi Informasi Universitas Lambung Mangkurat (JTIULM), vol. 6, no. 1, pp. 25–30, 2021, doi:
10.20527/jtiulm.v6i1.62.
[13] H. Prasetyo and B. A. Putra Akardihas, “Batik image retrieval using convolutional neural network,” Telkomnika
(Telecommunication Computing Electronics and Control), vol. 17, no. 6, pp. 3010–3018, 2019, doi:
10.12928/TELKOMNIKA.v17i6.12701.
[14] J. Tristanto, J. Hendryli, and D. Erny Herwindiati, “Classification of batik motifs using convolutional neural networks,” SSRN
Electronic Journal, 2018, doi: 10.2139/ssrn.3258935.
[15] Y. Gultom, A. M. Arymurthy, and R. J. Masikome, “Batik classification using deep convolutional network transfer learning,”
Jurnal Ilmu Komputer dan Informasi, vol. 11, no. 2, p. 59, 2018, doi: 10.21609/jiki.v11i2.507.
[16] N. D. Girsang, “Literature study of convolutional neural network algorithm for batik classification,” Brilliance: Research of
Artificial Intelligence, vol. 1, no. 1, pp. 1–7, 2021, doi: 10.47709/brilliance.v1i1.1069.
[17] A. E. Minarno, M. C. Mustaqim, Y. Azhar, W. A. Kusuma, and Y. Munarko, “Deep convolutional generative adversarial network
application in batik pattern generator,” 2021 9th International Conference on Information and Communication Technology,
ICoICT 2021, pp. 54–59, 2021, doi: 10.1109/ICoICT52021.2021.9527514.
[18] H. Fonda, “Classification of Riau batik using convolutional neural networks (CNN)" (in Indonesian) "Klasifikasi batik riau
dengan menggunakan convolutional neural networks (CNN),” Jurnal Ilmu Komputer, vol. 9, no. 1, pp. 7–10, 2020, doi:
10.33060/jik/2020/vol9.iss1.144.
[19] D. G. T. Meranggi, N. Yudistira, and Y. A. Sari, “Batik classification using convolutional neural network with data
improvements,” International Journal on Informatics Visualization, vol. 6, no. 1, pp. 6–11, 2022, doi: 10.30630/joiv.6.1.716.
[20] N. W. Parwati Septiani et al., “Convolutional neural network (CNN) algorithm for geometrical Batik Sade’ motifs,” ICCoSITE
2023 - International Conference on Computer Science, Information Technology and Engineering: Digital Transformation
Strategy in Facing the VUCA and TUNA Era, pp. 597–602, 2023, doi: 10.1109/ICCoSITE57641.2023.10127829.
[21] Z. Widyantoko, T. P. Widowati, Isnaini, and P. Trapsiladi, “Expert role in image classification using cnn for hard to identify
object: Distinguishing batik and its imitation,” IAES International Journal of Artificial Intelligence, vol. 10, no. 1, pp. 93–100,
2021, doi: 10.11591/ijai.v10.i1.pp93-100.
[22] C. U. Khasanah, E. Utami, and S. Raharjo, “Implementation of data augmentation using convolutional neural network for Batik
classification,” 2020 8th International Conference on Cyber and IT Service Management, CITSM 2020, 2020, doi:
10.1109/CITSM50537.2020.9268890.
[23] Y. Azhar, M. C. Mustaqim, and A. E. Minarno, “Ensemble convolutional neural network for robust batik classification,” IOP
Conference Series: Materials Science and Engineering, vol. 1077, no. 1, p. 012053, 2021, doi: 10.1088/1757-
899x/1077/1/012053.
[24] I. Nurhaida, V. Ayumi, D. Fitrianah, R. A. M. Zen, H. Noprisson, and H. Wei, “Implementation of deep neural networks (DNN)
with batch normalization for batik pattern recognition,” International Journal of Electrical and Computer Engineering, vol. 10,
no. 2, pp. 2045–2053, 2020, doi: 10.11591/ijece.v10i2.pp2045-2053.
[25] D. A. Mardani, Pranowo, and A. J. Santoso, “Deep learning for recognition of Javanese Batik patterns,” AIP Conference
Proceedings, vol. 2217, 2020, doi: 10.1063/5.0000686.
[26] M. A. Rasyidi and T. Bariyah, “Batik pattern recognition using convolutional neural network,” Bulletin of Electrical Engineering
and Informatics, vol. 9, no. 4, pp. 1430–1437, 2020, doi: 10.11591/eei.v9i4.2385.

271
[27] A. E. Minarno, F. D. S. Sumadi, H. Wibowo, and Y. Munarko, “Classification of batik patterns using K-nearest neighbor and
support vector machine,” Bulletin of Electrical Engineering and Informatics, vol. 9, no. 3, pp. 1260–1267, 2020, doi:
10.11591/eei.v9i3.1971.
[28] B. J. Filia et al., “Improving batik pattern classification using CNN with advanced augmentation and oversampling on imbalanced
dataset,” Procedia Computer Science, vol. 227, pp. 508–517, 2023, doi: 10.1016/j.procs.2023.10.552.
BIOGRAPHIES OF AUTHORS
Rosalina a respected lecturer in the Informatics Study Program at President
University, exemplifies dedication to her field. Her commitment to academic excellence is
underscored by her Master s degree in Informatics from President University. With a strong
educational background and a passion for informatics, Rosalina has a significant impact on
student’s academic journeys. Her expertise in the subject matter, combined with her ability to
explain complex concepts, creates a dynamic and enriching learning environment. Those
wishing to contact Rosalina can reach her via email at rosalina@president.ac.id
Genta Sahuri a dedicated lecturer at President University s Information Systems
Study Program, brings a wealth of knowledge and expertise to his role. Holding a Master s
degree in Informatics from the same institution underscores his commitment to academic
excellence. With a strong educational background and a passion for his field, Genta plays a
crucial role in shaping the academic journey of his students. His adeptness in conveying
complex concepts fosters a dynamic and enriching learning environment. He dedication to his
field and his ability to inspire students make him a valuable asset to President University. He
can be contacted at email: genta.sahuri@president.ac.id.

Unraveling Indonesian heritage through pattern recognition using YOLOv5

More Related Content

More from CSITiaesprime (20)

Recently uploaded (20)

Unraveling Indonesian heritage through pattern recognition using YOLOv5