Article Per Year (5 Year)
p-Index From 2021 - 2026
0.702
P-Index
This Author published in this journals
All Journal Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi) Jurnal Komputer dan Teknologi (JUKOMTEK) Jurnal Sistem Informasi dan Aplikasi
Willy Muhammad Fauzi
Politeknik Negeri Subang
Computer Science & IT Decision Sciences, Operations Research & Management Engineering Library & Information Science

Published : 4 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 4 Documents
Search

 1 
Attention-Based Multi-View Fusion YOLO for Non-Destructive Pineapple Sweetness Assessment Dwi Vernanda; Tri Herdiawan Apandi; Aisyah Suhailla Binti Jili; Desy Triastuti; Willy Muhammad Fauzi
Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi) Vol 10 No 2 (2026): April 2026
Publisher : Ikatan Ahli Informatika Indonesia (IAII)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29207/resti.v10i2.7382

Abstract

Classifying the sweetness level of pineapples is an important part of quality control, but existing methods still face issues of subjectivity and require destructive testing. Manual assessment is often inconsistent, while refractometer measurements require cutting the fruit open. Single-view computer vision offers a non-destructive alternative, yet its performance remains limited because visual cues related to sweetness appear on different sides of the fruit. This study introduces Multi-View Fusion YOLO (MVF-YOLO), a model that combines five viewpoints (full, front, left, right, and back) through an attention mechanism to perform adaptive sweetness classification. The dataset consists of 570 pineapples with TSS/TA ratios as the ground truth, producing 2,850 images grouped into three categories: sour (TSS/TA 10–20), ideal (TSS/TA 20–30), and very sweet (TSS/TA >30). MVF-YOLO achieved an mAP@0.5 of 82.1% and an overall accuracy of 84.2%, outperforming the single-view baseline by 14.3%. Attention weight analysis indicates that the full view contributes the most (0.267). With an inference time of 45.8 ms per fruit, the model is sufficiently efficient for use by farmers, distributors, and consumers. The results demonstrate that a multiview approach enhanced with learned attention can significantly improve sweetness classification accuracy without compromising computational efficiency.
EVALUASI KINERJA MODEL YOLOv11 UNTUK KLASIFIKASI TINGKAT KEMANISAN BUAH NANAS Willy Muhammad Fauzi; Dwi Vernanda; Tri Herdiawan Apandi
Jurnal Komputer dan Teknologi Vol 5 No 2 (2026): JUKOMTEK JULI 2026
Publisher : Yayasan Pendidikan Cahaya Budaya Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64626/jukomtek.v5i2.734

Abstract

Sweetness is one of the main determinants of pineapple quality, yet its conventional measurement through the TSS/TA ratio requires cutting the fruit open, making it unsuitable for non-destructive, large-scale sorting. Prior work by our group has explored non-destructive classification approaches, including a multi-view attention-based fusion model. As part of the iterative model development process toward that solution, this paper reports and analyzes the performance of a simpler baseline: a single-view YOLO object detection model trained directly to localize and classify pineapples into three sweetness categories-Asam (sour), Manis Ideal (ideal), and Sangat Manis (very sweet)-from a single RGB image per fruit. The model was trained for 50 epochs and evaluated using standard object detection metrics. The baseline achieved an overall mAP@0.5 of 0.555 and mAP@0.5:0.95 of 0.460, with the best F1-score of 0.58 reached at a confidence threshold of 0.183. Per-class analysis shows that the Asam category was the easiest to detect (mAP@0.5 = 0.695), while Manis Ideal (0.505) and Sangat Manis (0.465) were considerably weaker. Confusion matrix analysis at the default confidence threshold reveals that only 32-64% of ground-truth instances per class were correctly classified, notably lower than the recall trend suggested during training, and that the Sangat Manis class-the smallest in the dataset-was most frequently confused with its visual neighbor, Manis Ideal. These findings indicate that a single viewpoint, without any imbalance handling, is not yet sufficient to reliably separate boundary categories, providing empirical grounds for the multi-view and attention-based refinements explored in the continuation of this research.
Klasifikasi Jenis Beras Menggunakan Deep Learning Berbasis Computer Vision dengan Platform Roboflow Imron Hefni F; Muhamad Soleh Sulaeman; Tri Herdiawan Apandi; Willy Muhammad Fauzi
Jurnal Sistem Informasi dan Aplikasi (JSIA) Vol 4 No 1 (2026): Maret: Sistem Informasi
Publisher : Fakultas Ilmu Komputer

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.52958/jsia.v4i1.13175

Abstract

Rice is a fundamental food commodity globally, where accurate variety classification is crucial for pricing, quality control, and food security. Manual classification methods are labor-intensive, time-consuming, and prone to human subjectivity. This research proposes an automated classification system for five rice varieties (Arborio, Basmati, Ipsala, Jasmine, and Karacadag) using a Computer Vision approach with Vision Transformer (ViT) architecture. Unlike Convolutional Neural Networks (CNN) which focus on local features, ViT utilizes selfattention mechanisms to capture global contextual relationships within images. The model was developed using the public "Rice Image Dataset" containing 75,000 images. The methodology includes image preprocessing (resizing and normalization) and training of the ViT Classification model. Model performance was evaluated using standard metrics on a separate test set. The results show that the proposed ViT model achieved an outstanding accuracy of 99.9%. Thesefindings demonstrate that the Transformer-based approach is highly effective and efficient for automating rice variety identification,offering a more robust solution compared to conventional methods.
Penerapan Arsitektur Vision Transformer (ViT) Berbasis Low-Code untuk Klasifikasi Penyakit Padi Lucky Jayadinata; Giga Lukman Maulana; Tri Herdiawan Apandi; Willy Muhammad Fauzi
Jurnal Sistem Informasi dan Aplikasi (JSIA) Vol 4 No 1 (2026): Maret: Sistem Informasi
Publisher : Fakultas Ilmu Komputer

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.52958/jsia.v4i1.13176

Abstract

Rice plant diseases can significantly reduce productivity and threaten food security, making early and accurate detection essential. Traditional manual inspection methods are slow, subjective, and difficult to scale for field conditions. This study aims to develop a rice disease classification system using computer vision and the low-code Roboflow platform with the Vision Transformer (ViT) architecture. The Rice Diseases Image Dataset, consisting of four classes—BrownSpot, Healthy, Hispa, and LeafBlast—underwent preprocessing, data augmentation, and splitting before being trained using the ViT model. The best-performing model (Version 6) achieved an accuracy of 92,2% on the test set. The results demonstrate that the proposed low-code workflow effectively streamlines the deployment pipeline of rice disease classification models, achieving competitive performance suitable for rapid prototyping in precision agriculture. These findings provide an initial foundation for the application of computer vision technologies in supporting precision agriculture practices.
Co-Authors Aisyah Suhailla Binti Jili Desy Triastuti Dwi Vernanda Giga Lukman Maulana Imron Hefni F Lucky Jayadinata Muhamad Soleh Sulaeman Tri Herdiawan Apandi