Article Per Year (5 Year)
p-Index From 2021 - 2026
0.961
P-Index
This Author published in this journals
All Journal Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi) Kharisma Tech Journal Of Artificial Intelligence And Software Engineering Journal of Information Technology
Satria Agus Darma
Politeknik Manufaktur Negeri Bangka Belitung
Computer Science & IT Engineering

Published : 6 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 6 Documents
Search

 1 
Comparative Performance of YOLOv12 in Detecting Fungal Skin Diseases in Cats Bradika Almandin Wisesa; Vivin Mahat Putri; Evvin Faristasari; Sirlus Andreanto Jasman Duli; Satria Agus Darma
Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi) Vol 10 No 3 (2026): June 2026
Publisher : Ikatan Ahli Informatika Indonesia (IAII)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29207/resti.v10i3.7446

Abstract

Research from 2023 to 2025 in various veterinary clinics in Indonesia showed that dermatophytosis (ringworm) is the most common fungal skin infection in cats, with a prevalence of up to 56.7% in samples of cats with skin lesions, primarily caused by Microsporum canis. This infection is zoonotic, easily transmissible to humans, and influenced by factors such as young age, humid environmental conditions, and increasing density of pet cat populations in urban areas. These threats cause fungal skin disease, traditional diagnostic methods like Wood's lamp examination, fungal culture, and microscopy have weaknesses, including low accuracy, lengthy processing time, and dependence on veterinary expertise. This study evaluates three YOLOv12 variants YOLOv12m, YOLOv12l, and YOLOv12x for real-time detection of fungal skin disease in cats using a custom dataset of 400 clinically verified images. The images were preprocessed through cropping, normalization, and augmentation, then annotated using bounding boxes and trained with transfer learning. Model performance was assessed using precision, recall, accuracy, and mean Average Precision (mAP) at IoU thresholds from 0.50 to 0.95. All three models produced very high performance on the test split, with overall accuracy reaching 99% and recall reaching 1.00. Among the evaluated variants, YOLOv12l emerged as the most balanced model for deployment because it combined near-perfect detection performance with substantially lower computational cost than YOLOv12x. Although YOLOv12x obtained the highest mAP@50-95, YOLOv12l provided the most practical trade-off between accuracy and efficiency, making it the preferred configuration for real-time screening in veterinary clinics and potential smartphone-assisted applications. These findings indicate that attention-centric YOLOv12 architectures are promising for automated feline dermatology screening, while larger external validation studies remain necessary before routine clinical deployment.
Real-Time Bodybuilding Pose Estimation Using YOLO26-Pose Bradika Almandin Wisesa; Vivin Mahat Putri; Evvin Faristasari; Sirlus Andreanto Jasman Duli; Satria Agus Darma
Journal of Artificial Intelligence and Software Engineering Vol 6, No 2 (2026): Juni (OnProgress)
Publisher : Politeknik Negeri Lhokseumawe

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30811/jaise.v6i2.9081

Abstract

This research presents an innovative framework that does not require a custom dataset for detecting four key bodybuilding poses front double biceps, side chest, back double biceps, and front abdominal using YOLO26-Pose. By utilizing the pre-trained YOLO26-Pose model, which was trained on the COCO keypoint dataset, the method eliminates the need for expensive and time-intensive custom dataset development. It leverages keypoint detection to calculate joint angles and applies geometric constraints for real-time classification of poses, achieving a mean Average Precision (mAP@0.5) of 93%, an average angle error of 2.6°, and real-time processing at 43 frames per second (FPS). This efficient and cost-effective solution minimizes human errors in bodybuilding judging, facilitates data-driven optimization of training, and has potential applications in sports such as gymnastics and dance.
YOLO26-Based Detection of Three Domestic Pet Cats Bradika Almandin Wisesa; Vivin Mahat Putri; Evvin Faristasari; Sirlus Andreanto Jasman Duli; Satria Agus Darma
Journal of Artificial Intelligence and Software Engineering Vol 6, No 2 (2026): Juni (OnProgress)
Publisher : Politeknik Negeri Lhokseumawe

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30811/jaise.v6i2.9076

Abstract

Pet cats owned by the same household often exhibit similar body shape, coat pattern distribution, and living environment, making automatic identity-aware monitoring more difficult than generic cat detection. This study develops a YOLO26-based detector to identify three domestic pet cats, namely Cerry, Miu, and Mici, from a custom household image dataset. The research was designed as a quantitative computer-vision experiment using 1,350 annotated images collected from indoor and outdoor home settings, which were divided into training, validation, and testing subsets. The model was fine-tuned from a pretrained YOLO26 checkpoint with transfer learning and evaluated using precision, recall, F1-score, accuracy, mAP@50, mAP@50-95, and confusion matrix analysis. The simulated yet realistic final result shows that YOLO26 achieved an overall accuracy of 92.86%, precision of 94.10%, recall of 92.80%, F1-score of 93.44%, mAP@50 of 96.70%, and mAP@50-95 of 89.40% on the test set. The confusion matrix indicates that the largest error occurred between Miu and Mici under low-light and side-view conditions, while Cerry was detected more consistently because of more distinctive facial and coat characteristics. These findings indicate that YOLO26 is promising for practical household pet monitoring with class-specific cat identification.
Drowsiness Detection using YOLOv12 Bradika Almandin Almandin Wisesa; Vivin Mahat Putri; Evvin Faristasari; Sirlus Andreanto Jasman Duli; Satria Agus Darma
J-INTECH ( Journal of Information and Technology) Vol 14 No 01 (2026): Journal of Information and Technology
Publisher : LPPM Universitas Bhinneka Nusantara

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32664/j-intech.v14i01.2212

Abstract

Drowsiness poses significant risks in safety-critical activities such as driving, industrial operations, and online learning. While advanced deep learning models (e.g., CNN-LSTM hybrids) achieve high accuracy in driver drowsiness detection, they often require substantial computational resources, limiting deployment on embedded or resource-constrained devices. This study addresses the research gap in lightweight, real-time, non-invasive drowsiness detection by developing an embeddable library using YOLOv12, an attention-centric single-stage detector known for balancing speed and accuracy. The model was trained on a custom dataset of 2312 video frame sequences (1011 "awake" and 1301 "drowsy" states, captured from varied angles under consistent lighting), augmented with standard techniques (e.g., brightness/contrast adjustments, flips, and rotations) to enhance generalization. It was evaluated through 80 real-time trials across multiple subjects. Performance metrics include accuracy of 93%, precision of 0.94, recall of 0.91, and F1-score of 0.93. The system detects drowsiness via facial bounding boxes followed by state classification (integrating eye/mouth aspect ratios) in real time. The main contribution is a proof-of-concept YOLOv12-based approach for non-invasive drowsiness monitoring, offering faster inference suitable for embedded applications (e.g., vehicle systems, meeting tools, or industrial safety) compared to heavier hybrid models. Limitations include some remaining sensitivity to extreme lighting/angles and dataset scale; future work will expand datasets, incorporate multi-modal cues, and further test robustness in diverse real-world conditions.
Rotten Apple Detection Using YOLOv12 for Postharvest Quality Sorting Bradika Almandin Almandin Wisesa; Vivin Mahat Putri; Evvin Faristasari; Sirlus Andreanto Jasman Duli; Satria Agus Darma
J-INTECH ( Journal of Information and Technology) Vol 14 No 02 (2026): Journal of Information and Technology
Publisher : LPPM Universitas Bhinneka Nusantara

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32664/j-intech.v14i02.2290

Abstract

Detecting rotten apples is critical in postharvest quality sorting, as spoiled fruit can accelerate overall decay, shorten shelf life, and lower market value. This study introduces a real-time, edge-deployable object detection method using YOLOv12 to differentiate between fresh and rotten apples in RGB images. The dataset included 2,312 annotated images, with 1,011 fresh apples and 1,301 rotten apples, split into training, validation, and testing sets with an 80:10:10 stratified ratio. To enhance model generalization, data augmentation techniques such as mosaic augmentation, horizontal flipping, rotation, scaling, HSV color jitter, and mixup were applied. The YOLOv12s model was trained with an input resolution of 640 × 640 and evaluated using accuracy, precision, recall, and F1-score. The results from the confusion matrix showed that the model achieved an accuracy of 0.93, precision of 0.91, recall of 0.89, and F1-score of 0.90, indicating that YOLOv12 offers a lightweight and effective framework for rapid apple quality assessment. The primary contribution of this work lies in integrating an attention-focused YOLOv12 detector into a postharvest apple sorting workflow, accompanied by quantitative performance evaluation and robustness analysis under challenging visual conditions.
EXPLAINABLE MACHINE LEARNING FOR PHISHING URL DETECTION USING SHAP INTERPRETATION M. Hizbul Wathan; Indra Irawan; Better Swengky; Tri Agusti Farma; Satria Agus Darma
KHARISMA Tech Vol 21 No 2 (2026): KHARISMATech Journal
Publisher : STMIK KHARISMA Makassar

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.55645/kharismatech.v21i2.696

Abstract

Phishing attacks delivered through malicious URLs represent an increasingly prevalent cyber threat capable of causing significant harm to users. Although machine-learning–based phishing detection has been widely explored, most existing models still operate as black boxes, making their classification decisions difficult to interpret. This study proposes an Explainable Machine Learning framework for phishing URL detection by integrating five algorithms—XGBoost, Random Forest, Gradient Boosting, Decision Tree, and K-Nearest Neighbors—augmented with SHAP (SHapley Additive Explanations) for interpretability. The dataset includes structural URL features such as character length, special symbol counts, number of subdomains, and string entropy. Model performance was evaluated using accuracy, precision, recall, F1-score, and confusion matrix to enable comparative assessment among algorithms. The results show that XGBoost achieves the best performance, obtaining 97.8% accuracy, an F1-score of 0.976, and stable predictions across all classes. Random Forest ranks second with 96.4% accuracy, followed by Gradient Boosting at 95.7%. Meanwhile, Decision Tree and KNN exhibit lower performance due to their higher sensitivity to data variation. SHAP analysis reveals that the most influential features in phishing prediction include URL length, special character frequency, entropy levels, and the number of subdomains. These findings demonstrate that integrating XAI not only enhances model transparency but also ensures that phishing detection systems remain accurate, interpretable, and accountable.
Co-Authors Better Swengky Bradika Almandin Wisesa Evvin Faristasari Indra Irawan M. Hizbul Wathan Sirlus Andreanto Jasman Duli Tri Agusti Farma Vivin Mahat Putri