Implementasi YOLO Untuk Pengecekan Kelengkapan Pada Kotak Saji Pada Industri Katering

Ramadhan, Ditya Wahyu (2026) Implementasi YOLO Untuk Pengecekan Kelengkapan Pada Kotak Saji Pada Industri Katering. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5027221051-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Penelitian ini mengembangkan sistem quality control pengemasan makanan berbasis computer vision menggunakan algoritma YOLO untuk mendeteksi kelengkapan isi kotak saji secara real-time guna mengurangi kesalahan pada proses pengecekan di UMKM Culisnary. Sistem terdiri dari dua model utama, yaitu deteksi kotak saji dan deteksi isi makanan (nasi, lauk, pelengkap, sayur, dan buah) yang dilatih menggunakan dataset hasil kombinasi data manual dan Roboflow serta teknik augmentasi untuk meningkatkan generalisasi model. Evaluasi dilakukan menggunakan metrik mAP, precision, recall, pengujian real-time, serta uji robustness terhadap variasi pencahayaan, blur, dan jarak. Selain itu, dilakukan perbandingan beberapa model YOLO (v8n, v8s, v11n, v11s) serta benchmark perangkat keras Raspberry Pi 3 B+ dan Steam Deck untuk menguji performa inferensi. Hasil penelitian menunjukkan bahwa YOLOv8n memberikan keseimbangan terbaik antara akurasi, kecepatan, dan stabilitas sehingga digunakan pada implementasi sistem FoodQC Pro berbasis web dengan dashboard monitoring, dan hasil uji pengguna menunjukkan sistem mudah digunakan serta membantu proses quality control secara lebih efisien.
====================================================================================================================================================
This research develops a computer vision-based quality control system for food packaging using the YOLO (You Only Look Once) algorithm to automatically detect the completeness of food trays in real-time, aiming to reduce miss counting in UMKM (small and medium food industry) operations. The proposed system consists of two main models: a tray detection model and a food item detection model, including rice, side dishes, vegetables, and fruits. The models are trained using a combination of manually collected datasets and Roboflow datasets, along with data augmentation techniques to improve generalization under dynamic environmental conditions.Model evaluation is conducted using mean Average Precision (mAP), precision, recall, and real-time video testing. In addition, robustness testing is performed under varying conditions such as lighting changes, blur levels, and distance variations to assess model stability in real-world scenarios. A comparative study is also carried out between YOLOv8n, YOLOv8s, YOLOv11n, and YOLOv11s, as well as a hardware benchmark between Raspberry Pi 3 B+ and Steam Deck to evaluate inference performance. The results show that YOLOv8n provides the best balance between accuracy, speed, and stability, making it the most suitable model for implementation. The system is deployed as a web-based application called FoodQC Pro, equipped with a monitoring dashboard for real-time detection results. User evaluation from Culisnary UMKM indicates that the system is easy to use and effectively supports the food packaging quality control process.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Computer Vision, YOLO, Quality Control, Object Detection, Real-time Detection, UMKM, Food Packaging, Edge Computing
Subjects:	T Technology > T Technology (General) T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information Technology > 59201-(S1) Undergraduate Thesis
Depositing User:	Ditya Wahyu Ramadhan
Date Deposited:	28 Jul 2026 03:37
Last Modified:	28 Jul 2026 03:37
URI:	http://repository.its.ac.id/id/eprint/138376

Actions (login required)

View Item