I. Putu Agung Bayupati
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
I. Putu Agung Bayupati has a total of 54 co-authors. Here are the top co-authors by number of shared publications:
- I Made Agus Dwi Suarjaya
Universitas Udayana
4 shared publications - I. Made Sukarsa
Department of Information Technology, Faculty of Engineering, Udayana University, Bali, Indonesia
3 shared publications - Putu Wira Buana
Universitas Udayana
3 shared publications - Anak Agung Ngurah Hary Susila
Universitas Udayana
2 shared publications Gusti Agung Ayu Putri
Universitas Udayana, Badung
2 shared publications- I Ketut Gede Dharma Putra
Udayana University
2 shared publications
Latest Publications
- 8 April 2026
- Published by MECS Publisher in International Journal of Information Technology and Computer Science
- Vol. 18  (2) , 183-201
Accident Detection and Estimation of Vehicle Speed and Count by Type in Road CCTV Images Using Machine Vision
This study presents an integrated traffic monitoring system for accident detection, vehicle counting by type, and vehicle speed estimation using roadside Closed-Circuit Television (CCTV) footage and machine vision based on the YOLOv11 architecture. The proposed methodology comprises data collection from heterogeneous sources, data preprocessing and augmentation, model fine-tuning on a custom Vehicle–Accident dataset, system deployment through a web-based application, and real-world evaluation. The YOLOv11 models were optimized to detect multiple vehicle categories and clearly defined accident classes under real traffic conditions. Experimental results indicate that the YOLOv11 Large (l) model achieves superior detection performance, with 81.8% precision, 75.8% recall, 82.1% mAP50, and 53.3% mAP50–95. Real-world testing further confirms its effectiveness, yielding an object detection accuracy of 99.24% and low speed estimation errors, with Mean Absolute Percentage Error (MAPE) of 3.56% for video-based evaluation and 5.54% for real-time evaluation. In contrast, the YOLOv11 Nano (n) model offers faster inference and lower computational requirements but exhibits reduced robustness in complex accident scenarios. The trained models are deployed in an interactive web application supporting image, video, and real-time inputs, enabling practical traffic monitoring and decision support. Overall, the YOLOv11l-Vehicle-Accident model is identified as the most suitable configuration for accuracy-critical traffic management systems, while Nano variants are better suited for resource-constrained deployments.
- 8 April 2026
- Published by MECS Publisher in International Journal of Information Technology and Computer Science
- Vol. 18  (2) , 183-201
Accident Detection and Estimation of Vehicle Speed and Count by Type in Road CCTV Images Using Machine Vision
This study presents an integrated traffic monitoring system for accident detection, vehicle counting by type, and vehicle speed estimation using roadside Closed-Circuit Television (CCTV) footage and machine vision based on the YOLOv11 architecture. The proposed methodology comprises data collection from heterogeneous sources, data preprocessing and augmentation, model fine-tuning on a custom Vehicle–Accident dataset, system deployment through a web-based application, and real-world evaluation. The YOLOv11 models were optimized to detect multiple vehicle categories and clearly defined accident classes under real traffic conditions. Experimental results indicate that the YOLOv11 Large (l) model achieves superior detection performance, with 81.8% precision, 75.8% recall, 82.1% mAP50, and 53.3% mAP50–95. Real-world testing further confirms its effectiveness, yielding an object detection accuracy of 99.24% and low speed estimation errors, with Mean Absolute Percentage Error (MAPE) of 3.56% for video-based evaluation and 5.54% for real-time evaluation. In contrast, the YOLOv11 Nano (n) model offers faster inference and lower computational requirements but exhibits reduced robustness in complex accident scenarios. The trained models are deployed in an interactive web application supporting image, video, and real-time inputs, enabling practical traffic monitoring and decision support. Overall, the YOLOv11l-Vehicle-Accident model is identified as the most suitable configuration for accuracy-critical traffic management systems, while Nano variants are better suited for resource-constrained deployments.
- Open Access
- 12 September 2025
- Published by MDPI AG in Engineering Proceedings
- Vol. 107  (1) , 86
N-Gram and Full-Text Search Algorithm Testing for Pattern Recognition in a Chatbot Engine
The development of chatbots to access database services and information systems has triggered a lot of research on frameworks for service development, including the development of ISONER (Information System On Internet Messenger). This framework consists of multiple phases including pattern recognition, query processing, and response generation. In its implementation, the framework develops pattern recognition services that are currently based on Natural Language Processing (NLP). Improved pattern recognition algorithms enhance the system’s ability to accurately interpret user intent. The pattern recognition used in this research utilizes built-in plugins from MySQL, namely N-gram and Full-Text Search, which can be run directly on the MySQL engine to reduce latency and do not require another programming language. The FTS and fourgram algorithms gave the best results when applied on 100 test data points, with a threshold of 0.91, accuracy of 91%, precision of 99%, and recall of 92%; the average computation time was 19 s for 100 test data points and 2 min 49 s for 1000 data points tested simultaneously.
- 12 September 2025
- Published by MDPI AG in Engineering Proceedings
- Vol. 107  (1) , 86
Open AccessN-Gram and Full-Text Search Algorithm Testing for Pattern Recognition in a Chatbot Engine
The development of chatbots to access database services and information systems has triggered a lot of research on frameworks for service development, including the development of ISONER (Information System On Internet Messenger). This framework consists of multiple phases including pattern recognition, query processing, and response generation. In its implementation, the framework develops pattern recognition services that are currently based on Natural Language Processing (NLP). Improved pattern recognition algorithms enhance the system’s ability to accurately interpret user intent. The pattern recognition used in this research utilizes built-in plugins from MySQL, namely N-gram and Full-Text Search, which can be run directly on the MySQL engine to reduce latency and do not require another programming language. The FTS and fourgram algorithms gave the best results when applied on 100 test data points, with a threshold of 0.91, accuracy of 91%, precision of 99%, and recall of 92%; the average computation time was 19 s for 100 test data points and 2 min 49 s for 1000 data points tested simultaneously.
- Open Access
- 3 August 2025
- Published by Politeknik Negeri Batam in JOURNAL OF APPLIED INFORMATICS AND COMPUTING
- Vol. 9  (4) , 1858-1867
News Recommendation System Using Content-Based Filtering through RSS Customization Service
News refers to stories or information about current events or incidents. Several news websites offer a service called RSS (Really Simple Syndication), which enables users to easily receive updates on the latest news. News RSS feeds are typically generated based on the order of publication time or general categories. The content of these news RSS feeds can be customized to align with user interests or preferences. A recommendation system can be utilized as an approach to customize RSS feeds. This study was conducted to design a system capable of generating RSS feeds based on news recommendations using the content-based TF-IDF method and cosine similarity. Data scraping and preprocessing of news articles from various RSS feeds of Indonesian news websites were automated using cron jobs. Content-based filtering modeling was carried out using TF-IDF and cosine similarity. The design and customization of RSS feeds were implemented in a Flask application and packaged within several endpoints. The recommendations generated based on user click interactions were reasonably relevant, as they successfully presented news titles similar to the clicked articles, with cosine similarity scores ranging from 0.2 to 1.0. The majority of respondents agreed that the recommended news articles were relevant to the articles they had clicked and aligned with their interests. The RSS feed evaluation yielded highly satisfactory results, with all aspects assessed in the user acceptance survey achieving an average score exceeding 80%, and the overall results of the customer satisfaction survey indicated scores starting from 90%.
- 3 August 2025
- Published by Politeknik Negeri Batam in JOURNAL OF APPLIED INFORMATICS AND COMPUTING
- Vol. 9  (4) , 1858-1867
Open AccessNews Recommendation System Using Content-Based Filtering through RSS Customization Service
News refers to stories or information about current events or incidents. Several news websites offer a service called RSS (Really Simple Syndication), which enables users to easily receive updates on the latest news. News RSS feeds are typically generated based on the order of publication time or general categories. The content of these news RSS feeds can be customized to align with user interests or preferences. A recommendation system can be utilized as an approach to customize RSS feeds. This study was conducted to design a system capable of generating RSS feeds based on news recommendations using the content-based TF-IDF method and cosine similarity. Data scraping and preprocessing of news articles from various RSS feeds of Indonesian news websites were automated using cron jobs. Content-based filtering modeling was carried out using TF-IDF and cosine similarity. The design and customization of RSS feeds were implemented in a Flask application and packaged within several endpoints. The recommendations generated based on user click interactions were reasonably relevant, as they successfully presented news titles similar to the clicked articles, with cosine similarity scores ranging from 0.2 to 1.0. The majority of respondents agreed that the recommended news articles were relevant to the articles they had clicked and aligned with their interests. The RSS feed evaluation yielded highly satisfactory results, with all aspects assessed in the user acceptance survey achieving an average score exceeding 80%, and the overall results of the customer satisfaction survey indicated scores starting from 90%.
- Open Access
- 31 January 2025
- Published by Universitas Udayana in Lontar Komputer : Jurnal Ilmiah Teknologi Informasi
- Vol. 15  (03) , 149-160
Training VGG16, MobileNetV1 and Simple CNN Models from Scratch for Balinese Inscription Recognition
Many inscriptions in Bali are damaged. Damage to these inscriptions can be caused by natural disasters, overgrown with moss, algae and bacteria. Damage can also be caused by warfare, or deliberately erased. This inscription contains the knowledge and civilization of the ancestors so it is very important to be able to read its contents. Based on these problems, this research conducted training from scratch on 3 CNN models namely VGG16, MobileNetV1 and Simple CNN. The purpose of this research is to choose one recognition model that has the best performance and produces the highest recognition rate to proceed to the inscription restoration stage. The dataset used is Balinese inscription: Isolated Character Recognition of Balinese Script in Palm Leaf Manuscript Images in Challenge-3-ForTrain.zip. The training process of three models with five different training files resulted in the finding that VGG16 has the highest accuracy in the training, testing, and validation process with the least number of epochs. This research contributes to specific datasets, such as the Isolated Character Recognition of Balinese Script using the training process from the beginning of VGG16, involving all stages of the process. It will produce the best model performance compared to the other four training models.
- 31 January 2025
- Published by Universitas Udayana in Lontar Komputer : Jurnal Ilmiah Teknologi Informasi
- Vol. 15  (03) , 149-160
Open AccessTraining VGG16, MobileNetV1 and Simple CNN Models from Scratch for Balinese Inscription Recognition
Many inscriptions in Bali are damaged. Damage to these inscriptions can be caused by natural disasters, overgrown with moss, algae and bacteria. Damage can also be caused by warfare, or deliberately erased. This inscription contains the knowledge and civilization of the ancestors so it is very important to be able to read its contents. Based on these problems, this research conducted training from scratch on 3 CNN models namely VGG16, MobileNetV1 and Simple CNN. The purpose of this research is to choose one recognition model that has the best performance and produces the highest recognition rate to proceed to the inscription restoration stage. The dataset used is Balinese inscription: Isolated Character Recognition of Balinese Script in Palm Leaf Manuscript Images in Challenge-3-ForTrain.zip. The training process of three models with five different training files resulted in the finding that VGG16 has the highest accuracy in the training, testing, and validation process with the least number of epochs. This research contributes to specific datasets, such as the Isolated Character Recognition of Balinese Script using the training process from the beginning of VGG16, involving all stages of the process. It will produce the best model performance compared to the other four training models.
- Open Access
- 15 November 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16  (22) , 4264
Deep Learning Model Size Performance Evaluation for Lightning Whistler Detection on Arase Satellite Dataset
The plasmasphere within Earth’s magnetosphere plays a crucial role in space physics, with its electron density distribution being pivotal and strongly influenced by solar activity. Very Low Frequency (VLF) waves, including whistlers, provide valuable insights into this distribution, making the study of their propagation through the plasmasphere essential for predicting space weather impacts on various technologies. In this study, we evaluate the performance of different deep learning model sizes for lightning whistler detection using the YOLO (You Only Look Once) architecture. To achieve this, we transformed the entirety of raw data from the Arase (ERG) Satellite for August 2017 into 2736 images, which were then used to train the models. Our approach involves exposing the models to spectrogram diagrams—visual representations of the frequency content of signals—derived from the Arase Satellite’s WFC (WaveForm Capture) subsystem, with a focus on analyzing whistler-mode plasma waves. We experimented with various model sizes, adjusting epochs, and conducted performance analysis using a partial set of labeled data. The testing phase confirmed the effectiveness of the models, with YOLOv5n emerging as the optimal choice due to its compact size (3.7 MB) and impressive detection speed, making it suitable for resource-constrained applications. Despite challenges such as image quality and the detection of smaller whistlers, YOLOv5n demonstrated commendable accuracy in identifying scenarios with simple shapes, thereby contributing to a deeper understanding of whistlers’ impact on Earth’s magnetosphere and fulfilling the core objectives of this study.
- 15 November 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16  (22) , 4264
Open AccessDeep Learning Model Size Performance Evaluation for Lightning Whistler Detection on Arase Satellite Dataset
The plasmasphere within Earth’s magnetosphere plays a crucial role in space physics, with its electron density distribution being pivotal and strongly influenced by solar activity. Very Low Frequency (VLF) waves, including whistlers, provide valuable insights into this distribution, making the study of their propagation through the plasmasphere essential for predicting space weather impacts on various technologies. In this study, we evaluate the performance of different deep learning model sizes for lightning whistler detection using the YOLO (You Only Look Once) architecture. To achieve this, we transformed the entirety of raw data from the Arase (ERG) Satellite for August 2017 into 2736 images, which were then used to train the models. Our approach involves exposing the models to spectrogram diagrams—visual representations of the frequency content of signals—derived from the Arase Satellite’s WFC (WaveForm Capture) subsystem, with a focus on analyzing whistler-mode plasma waves. We experimented with various model sizes, adjusting epochs, and conducted performance analysis using a partial set of labeled data. The testing phase confirmed the effectiveness of the models, with YOLOv5n emerging as the optimal choice due to its compact size (3.7 MB) and impressive detection speed, making it suitable for resource-constrained applications. Despite challenges such as image quality and the detection of smaller whistlers, YOLOv5n demonstrated commendable accuracy in identifying scenarios with simple shapes, thereby contributing to a deeper understanding of whistlers’ impact on Earth’s magnetosphere and fulfilling the core objectives of this study.
- Open Access
- 11 May 2024
- Published by Atlantis Press SARL
- p. 17-28
A Study of Lab Color Space and Its Visualization
- 11 May 2024
- Published by Atlantis Press SARL
- p. 17-28
Open AccessA Study of Lab Color Space and Its Visualization
- Open Access
- 1 October 2023
- Published by Universitas Atma Jaya Yogyakarta in Jurnal Buana Informatika
- Vol. 14  (02) , 107-116
Analisis Sentimen Review Hotel Menggunakan Metode Deep Learning BERT
The COVID-19 pandemic has resulted in declining tourism visits and hotel occupancy. Hoteliers must monitor visitor lifestyles to sustain their businesses. One way to achieve this is by understanding the sentiment of hotel visitors through review analysis, enabling better decision-making regarding service and business aspects in the hotel industry. This research applies the natural language processing deep learning model BERT to analyze positive and negative sentiments from hotel visitor reviews in Indonesia. The BERT model undergoes a pre-trained and fine-tuned process to produce accurate sentiment analysis. Evaluation results demonstrate that the fine-tuned SmallBERT model performs well, trained on a dataset of 515k hotel reviews for five epochs. The SmallBERT model achieves an accuracy of 91.40%, precision of 90.51%, recall of 90.51%, and an F1 score of 90.51% when evaluated with manually labelled datasets. Visualizations of the predominantly positive sentiment comparisons are conducted using Tableau.
- 1 October 2023
- Published by Universitas Atma Jaya Yogyakarta in Jurnal Buana Informatika
- Vol. 14  (02) , 107-116
Open AccessAnalisis Sentimen Review Hotel Menggunakan Metode Deep Learning BERT
The COVID-19 pandemic has resulted in declining tourism visits and hotel occupancy. Hoteliers must monitor visitor lifestyles to sustain their businesses. One way to achieve this is by understanding the sentiment of hotel visitors through review analysis, enabling better decision-making regarding service and business aspects in the hotel industry. This research applies the natural language processing deep learning model BERT to analyze positive and negative sentiments from hotel visitor reviews in Indonesia. The BERT model undergoes a pre-trained and fine-tuned process to produce accurate sentiment analysis. Evaluation results demonstrate that the fine-tuned SmallBERT model performs well, trained on a dataset of 515k hotel reviews for five epochs. The SmallBERT model achieves an accuracy of 91.40%, precision of 90.51%, recall of 90.51%, and an F1 score of 90.51% when evaluated with manually labelled datasets. Visualizations of the predominantly positive sentiment comparisons are conducted using Tableau.
- Open Access
- 31 August 2023
- Published by Institut Bisnis dan Teknologi Indonesia in SINTECH (Science and Information Technology) Journal
- Vol. 6  (2) , 99-106
Pengembangan Sistem Business Intelligence Dalam Monitoring Performa Perusahaan Multi Company
Multi-company companies have challenges in managing subsidiary data into performance information of all subsidiaries in one window. This is due to the variation of data dimensions according to the business processes of each subsidiary. CV. XYZ is a holding company engaged in the food & beverage business. CV. XYZ currently manages three companies consisting of restaurants, catering, and tent and decoration rental. The problem faced by the company owner is the limited access to company performance in one information window. Each company has working papers in Excel format to record expenditure and income transactions. This article proposes the development of a website-based business intelligence system to overcome the problems of CV. XYZ. The purpose of developing a business intelligence system in this article is to provide access to the performance of each subsidiary in one website media. The business intelligence system is developed through the stages of data collection and analysis, data warehouse design, ETL process, and data visualization with Microsoft Power BI. The data warehouse design uses Kimball's nine-step method which produces a data warehouse with a star scheme. The developed Business Intelligence system was tested using the UAT method. The UAT test results show that the system development is following the company's needs as indicated by the UAT score of 92%
- 31 August 2023
- Published by Institut Bisnis dan Teknologi Indonesia in SINTECH (Science and Information Technology) Journal
- Vol. 6  (2) , 99-106
Open AccessPengembangan Sistem Business Intelligence Dalam Monitoring Performa Perusahaan Multi Company
Multi-company companies have challenges in managing subsidiary data into performance information of all subsidiaries in one window. This is due to the variation of data dimensions according to the business processes of each subsidiary. CV. XYZ is a holding company engaged in the food & beverage business. CV. XYZ currently manages three companies consisting of restaurants, catering, and tent and decoration rental. The problem faced by the company owner is the limited access to company performance in one information window. Each company has working papers in Excel format to record expenditure and income transactions. This article proposes the development of a website-based business intelligence system to overcome the problems of CV. XYZ. The purpose of developing a business intelligence system in this article is to provide access to the performance of each subsidiary in one website media. The business intelligence system is developed through the stages of data collection and analysis, data warehouse design, ETL process, and data visualization with Microsoft Power BI. The data warehouse design uses Kimball's nine-step method which produces a data warehouse with a star scheme. The developed Business Intelligence system was tested using the UAT method. The UAT test results show that the system development is following the company's needs as indicated by the UAT score of 92%
- Open Access
- 21 August 2023
- Published by Universitas Udayana in Jurnal Ilmiah Merpati (Menara Penelitian Akademika Teknologi Informasi)
- Vol. 11  (2) , 71-79
Design and Development of an Internet of Things Based Package Reception Box System
Parkage delivery services play a crucial role in facilitating the easy and prompt sending and receiving of goods to the public. The package delivery sector has witnessed significant growth of 33.62% since the outbreak of the Covid-19 pandemic. However, package receipt often presents challenges, including instances of damaged or lost packages. To address these issues, a system for package reception was designed and developed utilizing the Internet of Things (IoT) technology. The system operates by incorporating a UART GM66 Barcode Scanner, ESP-32CAM, Solenoid Door Lock, and Motor Stepper Nema 17. The primary objective of this research is to construct an automated package-reception system that can be controlled securely through a smartphone, thus ensuring protection against theft and package damage. System testing was performed on the functionality of the tool and communication with the telegram bot.
- 21 August 2023
- Published by Universitas Udayana in Jurnal Ilmiah Merpati (Menara Penelitian Akademika Teknologi Informasi)
- Vol. 11  (2) , 71-79
Open AccessDesign and Development of an Internet of Things Based Package Reception Box System
Parkage delivery services play a crucial role in facilitating the easy and prompt sending and receiving of goods to the public. The package delivery sector has witnessed significant growth of 33.62% since the outbreak of the Covid-19 pandemic. However, package receipt often presents challenges, including instances of damaged or lost packages. To address these issues, a system for package reception was designed and developed utilizing the Internet of Things (IoT) technology. The system operates by incorporating a UART GM66 Barcode Scanner, ESP-32CAM, Solenoid Door Lock, and Motor Stepper Nema 17. The primary objective of this research is to construct an automated package-reception system that can be controlled securely through a smartphone, thus ensuring protection against theft and package damage. System testing was performed on the functionality of the tool and communication with the telegram bot.
- Open Access
- 13 July 2023
- Published by Insight Society in International Journal on Advanced Science, Engineering and Information Technology
- Vol. 13  (4) , 1582-1588
Design and Build of Masked Face Identification System and IoT-Based Body Temperature Measurement
The new normal is an era in the behavior changed to obstruct the spread of COVID-19, such as decreasing people's mobility, body temperature measuring, mandatory masking, and getting a COVID-19 vaccine regularly. This study develops an identification system based on the Internet of Things through facial biometrics and temperature measurement. Face identification is divided into two main steps: face detection and identification. Face detections used the Framework YOLOv5, in which the systems can detect masked and without masked faces. Pre-trained VGG-face is used for face identification for feature extraction and produces a 2622-dimensional vector. The feature extraction result is calculated as the distance similarity with the features stored in the Database using Euclidean distance. Temperature measurement utilizes IoT by using the NodeMCU ESP8266 and the MLX90614 sensor. NodeMCU ESP8266 is a microcontroller equipped with a WI-FI module to send temperature data so measurements can be delivered wirelessly. The MLX90614 sensor measures body temperature at a 40 - 60 cm distance from the Sensor. Calibration of the sensor used Two-point Calibration, so a trim error rate level is produced. The result successfully identified the face with the F1 score of 92% without a masked face and 73% for a masked face. The body temperature was measured using the MLX90614 sensor produced an error rate of 0.1°C after calibration. In the future, this system can be further developed and utilized for other sectors, such as the medical and security sectors.
- 13 July 2023
- Published by Insight Society in International Journal on Advanced Science, Engineering and Information Technology
- Vol. 13  (4) , 1582-1588
Open AccessDesign and Build of Masked Face Identification System and IoT-Based Body Temperature Measurement
The new normal is an era in the behavior changed to obstruct the spread of COVID-19, such as decreasing people's mobility, body temperature measuring, mandatory masking, and getting a COVID-19 vaccine regularly. This study develops an identification system based on the Internet of Things through facial biometrics and temperature measurement. Face identification is divided into two main steps: face detection and identification. Face detections used the Framework YOLOv5, in which the systems can detect masked and without masked faces. Pre-trained VGG-face is used for face identification for feature extraction and produces a 2622-dimensional vector. The feature extraction result is calculated as the distance similarity with the features stored in the Database using Euclidean distance. Temperature measurement utilizes IoT by using the NodeMCU ESP8266 and the MLX90614 sensor. NodeMCU ESP8266 is a microcontroller equipped with a WI-FI module to send temperature data so measurements can be delivered wirelessly. The MLX90614 sensor measures body temperature at a 40 - 60 cm distance from the Sensor. Calibration of the sensor used Two-point Calibration, so a trim error rate level is produced. The result successfully identified the face with the F1 score of 92% without a masked face and 73% for a masked face. The body temperature was measured using the MLX90614 sensor produced an error rate of 0.1°C after calibration. In the future, this system can be further developed and utilized for other sectors, such as the medical and security sectors.