Deteksi Orientasi Tidak Normal Sosis Pada Konveyor Area Produksi Menggunakan YOLOv8 Dengan Metode Group-wise Rotating Attention

Alifiyan Rahmad Romadhoni, Alifiyan Rahmad Romadhoni (2026) Deteksi Orientasi Tidak Normal Sosis Pada Konveyor Area Produksi Menggunakan YOLOv8 Dengan Metode Group-wise Rotating Attention. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040221050-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Sebagai industri pengolahan sosis, PT Charoen Pokphand Indonesia Plant Berbek perlu menjaga konsistensi mutu produk, termasuk kesesuaian orientasi sosis selama proses produksi. Salah satu permasalahan yang masih ditemukan adalah orientasi sosis yang tidak normal pada konveyor produksi, yang berpotensi mengganggu proses pengemasan dan meningkatkan risiko produk cacat. Penelitian ini mengembangkan sistem deteksi orientasi tidak normal sosis berbasis Computer Vision menggunakan YOLOv8 yang dimodifikasi dengan metode Group-wise Rotating Attention (GRA) untuk meningkatkan kemampuan deteksi terhadap objek yang memiliki variasi orientasi. Pengujian menggunakan dataset original yang mencerminkan kondisi nyata produksi dan dataset skenario yang disusun lebih teratur. Hasil pengujian menunjukkan bahwa penerapan Group-wise Rotating Attention (GRA) pada YOLOv8 memberikan peningkatan performa dibandingkan model baseline pada dataset original, dengan nilai mAP50 sebesar 84,7%. Sementara itu, pada dataset skenario kedua model menunjukkan performa yang relatif setara. Evaluasi menggunakan Ground Truth terhadap 20 citra uji menghasilkan nilai Precision sebesar 95,23%, Recall sebesar 98,71%, F1-score sebesar 96,94%, dan Accuracy sebesar 94,06%, sehingga sistem mampu mendeteksi orientasi sosis sebagai pendukung proses inspeksi pada lingkungan produksi. Hasil penelitian menunjukkan bahwa integrasi metode GRA pada YOLOv8 mampu meningkatkan kemampuan deteksi orientasi sosis pada kondisi objek yang kompleks.
====================================================================================================================================
As a sausage processing industry, PT Charoen Pokphand Indonesia Plant Berbek is required to maintain consistent product quality, including ensuring the proper orientation of sausages during the production process. One of the problems encountered is the occurrence of abnormal sausage orientations on the production conveyor, which may disrupt the packaging process and increase the risk of defective products. This study develops a computer vision-based abnormal sausage orientation detection system using YOLOv8 modified with the Group-wise Rotating Attention (GRA) method to improve detection performance for objects with varying orientations. The model was evaluated using the original dataset, which represents actual production conditions, and the scenario dataset, which contains more structured object arrangements. The experimental results show that integrating GRA into YOLOv8 improves detection performance on the original dataset compared with the baseline model, achieving an mAP50 of 84.7%. Meanwhile, both models demonstrate comparable performance on the scenario dataset. Evaluation using Ground Truth on 20 test images achieved a Precision of 95.23%, Recall of 98.71%, F1-score of 96.94%, and Accuracy of 94.06%, indicating that the proposed system is capable of detecting sausage orientation effectively to support inspection processes in the production environment. Overall, the results demonstrate that integrating GRA into YOLOv8 enhances sausage orientation detection performance under complex object conditions.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	YOLOv8, Computer Vision, Convolutional Neural Network (CNN), Deteksi Orientasi Sosis, Group-wise Rotating Attention, YOLOv8, Computer Vision, Convolutional Neural Network (CNN), Sausages Orientation Detection, Group-wise Rotating Attention
Subjects:	Q Science Q Science > Q Science (General) Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Alifiyan Rahmad Romadhoni
Date Deposited:	10 Aug 2026 03:25
Last Modified:	10 Aug 2026 03:25
URI:	http://repository.its.ac.id/id/eprint/144280

Actions (login required)

View Item