Garuda - Garba Rujukan Digital

Optimizing YOLOv8 for Real-Time CCTV Surveillance: A Trade-off Between Speed and Accuracy Muhammad Rizqi Sholahuddin; Maisevli Harika; Iwan Awaludin; Yunita Citra Dewi; Fachri Dhia Fauzan; Bima Putra Sudimulya; Vandha Pradiyasma Widarta
JOIN (Jurnal Online Informatika) Vol 8 No 2 (2023)
Publisher : Department of Informatics, UIN Sunan Gunung Djati Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.15575/join.v8i2.1196

Real-time video surveillance, especially CCTV systems, requires fast and accurate face detection. Object detection models with slow inference times are ineffective in real-time. This study addresses this challenge by improving the inference speed of the YOLOv8 model, a leading object detection framework known for its accuracy and speed. We focus on pruning the model's architecture, particularly the P5 head section, which detects larger objects. According to Bochkovskiy's 2020 research, this modification enhances the model's performance specifically for medium and small objects in CCTV footage. The standard YOLOv8 model and its modified version were compared for inference time, mean Average Precision (mAP), and model weight. The pruned YOLOv8 model cuts inference time by 15.56%, from 4.5 ms to 3.8 ms, and reduces model weight. The advantages mentioned above are offset by a 1.6% decrease in mean average precision. This research advances object detection technology by demonstrating architectural modifications' efficacy. These changes make the model faster and lighter, making it suitable for real-time surveillance. The accuracy trade-off is slight. The implications of these findings are crucial for implementing efficient object detection systems in CCTV surveillance. These findings also lay the groundwork for future research to improve such systems' speed-accuracy trade-off.

Aplikasi Grafologi dari Huruf “t” Menggunakan Jaringan Syaraf Tiruan Iwan Awaludin; Aulia Khairunisa
Jurnal Nasional Teknik Elektro dan Teknologi Informasi Vol 4 No 3: Agustus 2015
Publisher : Departemen Teknik Elektro dan Teknologi Informasi, Fakultas Teknik, Universitas Gadjah Mada

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (215.575 KB)

Graphology is a branch of science which classifies human personality from handwriting. Graphologists observe the patterns of handwriting and compare it with personality class database. Computers can be trained to do the same procedure of human personality classification based on handwriting. The procedure is to perform digital image processing that extracts features from handwriting images. The features will become input for Artificial Neural Network. Neural networks that are already configured with a number of hidden layers, the number of neurons, activation function, and the particular learning algorithm will be able to recognize certain classes of human handwriting, thus his personality. Tested configurations include: changing the number of neurons in the hidden layer of eight to twelve, binary image resizing, changing the activation function, and also changing the learning algorithm. Results of simulation and analysis are also provided.

Optimizing YOLOv8 for Real-Time CCTV Surveillance: A Trade-off Between Speed and Accuracy Muhammad Rizqi Sholahuddin; Maisevli Harika; Iwan Awaludin; Yunita Citra Dewi; Fachri Dhia Fauzan; Bima Putra Sudimulya; Vandha Pradiyasma Widarta
JOIN (Jurnal Online Informatika) Vol 8 No 2 (2023)
Publisher : Department of Informatics, UIN Sunan Gunung Djati Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.15575/join.v8i2.1196

Real-time video surveillance, especially CCTV systems, requires fast and accurate face detection. Object detection models with slow inference times are ineffective in real-time. This study addresses this challenge by improving the inference speed of the YOLOv8 model, a leading object detection framework known for its accuracy and speed. We focus on pruning the model's architecture, particularly the P5 head section, which detects larger objects. According to Bochkovskiy's 2020 research, this modification enhances the model's performance specifically for medium and small objects in CCTV footage. The standard YOLOv8 model and its modified version were compared for inference time, mean Average Precision (mAP), and model weight. The pruned YOLOv8 model cuts inference time by 15.56%, from 4.5 ms to 3.8 ms, and reduces model weight. The advantages mentioned above are offset by a 1.6% decrease in mean average precision. This research advances object detection technology by demonstrating architectural modifications' efficacy. These changes make the model faster and lighter, making it suitable for real-time surveillance. The accuracy trade-off is slight. The implications of these findings are crucial for implementing efficient object detection systems in CCTV surveillance. These findings also lay the groundwork for future research to improve such systems' speed-accuracy trade-off.

Analisis Data Eksplorasi Klasifikasi Aktivitas Otak yang Berbahaya Putriadhinia, Salma Syawalan; Mulia, Syelvie Ira Ratna; Awaludin, Iwan; Sholahuddin, Muhammad Rizqi; Syakrani, Nurjannah; Hayati, Hashri
Prosiding Industrial Research Workshop and National Seminar Vol 15 No 1 (2024): Prosiding 15th Industrial Research Workshop and National Seminar (IRWNS)
Publisher : Politeknik Negeri Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35313/irwns.v15i1.6234

Elektroensefalografi (EEG) merupakan alat yang vital dalam rekaman dan analisis aktivitas listrik otak, sering digunakan dalam penelitian dan perawatan medis. Peletakan elektroda EEG mengikuti sistem internasional 10-20, dengan huruf dan angka tertentu untuk menandakan lokasi spesifik di otak. Kualitas pengukuran EEG sangat penting, dengan upaya mengeliminasi artifact yang bisa berasal dari sumber biologis maupun nonbiologis. Monitoring EEG di ICU telah meningkat, terutama untuk mendeteksi pola IIIC yang berbahaya. Pola tersebut sulit dibedakan dari kejang biasa dan dapat menyebabkan kerusakan otak. Penelitian ini bertujuan untuk melakukan analisis terhadap dataset EEG yang memiliki pola IIIC sehingga harapannya dapat berguna untuk peneliti yang hendak menggunakan data tersebut. Penelitian ini menggunakan dataset dari platform Kaggle, tepatnya HMS – Harmful Brain Activity Classification. Dataset tersebut memiliki data mentah EEG dan spektogram yang sudah dianotasi oleh ahli. Analisis data menunjukkan bahwa dataset tersebut memiliki keseimbangan jumlah data yang dianotasi untuk masing-masing kategori IIIC. Dalam dataset tersebut, terdapat data rekaman EEG dan data spektogram yang memiliki nilai kosong (null value) sehingga perlu dilakukan penangan terlebih dahulu sebelum diolah lebih lanjut.

Dataset Citra Papan Sirkuit Tercetak dengan Komponen yang Terbakar Awaludin, Iwan; Gelar, Trisna; Sholahuddin, Muhammad Rizqi; Melinia, Gina; Kadhafi, Irvan; Sitepu, Rezky Wahyuda
Building of Informatics, Technology and Science (BITS) Vol 3 No 3 (2021): December 2021
Publisher : Forum Kerjasama Pendidikan Tinggi

The application of artificial intelligence, especially in the automatic optical inspection of printed circuit boards or PCBs, is increasingly being carried out by researchers. Unfortunately, the data used to train and test artificial intelligence models is synthetic data. Printed circuit boards in good condition are imaged and then changed by software to give the impression of defects. In addition, the type of damage is limited to pre-operation, namely when the PCB is not yet operational. After the PCB is operational, damage can occur, for example, burned components. Until now, there is no data set of PCB images with burned components. This study, therefore, explores data retrieval techniques that can produce the required data set. This data collection technique includes hardware setup and PCB data sources. Based on the exploration results, it is concluded that a trinocular digital microscope with high resolution can produce sharp PCB images. The obstacle that arises is the difficulty of getting PCBs with burned components. The solution was obtained by referring to the PCB repair video from the Youtube channel. Several data were collected and tested with EfficientDet with 90% mAP.

COST AND TIME EFFICIENCY ANALYSIS OF ROAD CONDITION SURVEYS: MANUAL VS SEMI-AUTOMATED METHODS Sari, Rifdah Puspita; Astor, Yackob; Awaludin, Iwan; Shalahuddin, Salman
Jurnal Pensil : Pendidikan Teknik Sipil Vol. 14 No. 3 (2025): Jurnal Pensil : Pendidikan Teknik Sipil
Publisher : LPPM Universitas Negeri Jakarta

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21009/jpensil.v14i3.58676

Roads are a fundamental component of transportation that plays a critical role in national economic growth. Maintaining road conditions is essential to ensure optimal traffic serviceability. However, in developing countries like Indonesia, these surveys are predominantly conducted manually. This conventional approach is time-consuming, costly, and requires a substantial amount of human resources. The swift progression of machine learning (ML) within Artificial Intelligence (AI) presents an opportunity to be utilized as a data processing tool for road condition surveys, leading to greater time and cost efficiency. This study analyzes the cost and time required for two survey methods: manual and semi-automated, employing machine learning. Based on the analysis conducted on 164 km of urban roads in Bandung City, the semi-automated ML method achieved a cost efficiency of 72.23%, with its total cost being only 27.77% of the manual method. Furthermore, the time efficiency reached 96.34%, meaning the survey was completed in just 3.66% of the time required by the manual approach. These results indicate that the application of machine learning for semi-automated road condition surveys is substantially more efficient in terms of both time and cost compared to traditional manual surveys.

Iron Ball Launcher Platform Control System for Impact Test at Glass Testing Laboratory Iwan Awaludin; Muhammad Rizqi Sholahuddin; Yudi Widhiyasana; Sofy Fitriani
Jurnal Serambi Engineering Vol. 10 No. 4 (2025): Oktober 2025
Publisher : Faculty of Engineering, Universitas Serambi Mekkah

Show Abstract | Download Original | Original Source | Check in Google Scholar

This study presents the digital transformation of a glass impact test system at the Center for Standardization and Services for the Ceramic and Non-Metallic Mineral Industry (BBK) through Industry 4.0 integration. The legacy system faced challenges including manual distance measurement, outdated safety components, mechanical momentum causing positioning inaccuracy, and inability to sequentially launch multiple iron balls. To address these, a phased approach was implemented: analysis, design, implementation, and testing of a digital control system. Key upgrades included LIDAR-based wireless distance sensing (up to 9 meters), RS-485 communication for reliable data transfer, replacement of 1980s-era fuses with modern Mini Circuit Breakers, and algorithmic compensation for mechanical delay. A microcontroller-based control system enabled automated height adjustment, mode selection per national standards, and sequential ball release. The system was tested across six height settings with five trials each, achieving an error rate below 1% in all cases. Results confirm enhanced precision, safety, and efficiency. This targeted digitalization demonstrates how Industry 4.0 technologies can modernize legacy testing equipment without full replacement, offering a cost-effective, scalable model for industrial laboratories undergoing digital transformation.

Title

Found 17 Documents
Search

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Title Search

Found 17 Documents Search

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Title

Found 17 Documents
Search