Kurniawan, Muhammad Andi (2025) Sistem Deteksi Produk Sosis Bengkok Dan Normal Pada Sistem Sortir Menggunakan Metode Convolutional Neural Network (CNN). Diploma thesis, Institut Teknologi Sepuluh Nopember.
![]() |
Text
2040211056_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy |
Abstract
Penyortiran produk sosis bengkok secara manual di industri makanan sering mengalami kendala akibat subyektivitas penilaian operator serta faktor kelelahan, yang menyebabkan variasi dalam kualitas produk. Penelitian ini merancang sistem sortir otomatis berbasis Convolutional Neural Network (CNN) dengan menggunakan model YOLOv5 (s, m, l, x) dan YOLOv8 (s, m, l, x) untuk mendeteksi dan mengklasifikasikan sosis normal dan sosis bengkok. Dataset penelitian terdiri dari 1.461 citra sosis yang dikumpulkan dari conveyor prototipe dan conveyor produksi, kemudian melalui tahap anotasi dan augmentasi untuk meningkatkan variasi citra dalam pelatihan model. Pelatihan model dilakukan di Google Colaboratory dengan parameter evaluasi mencakup precision, recall, F1-Score, dan mAP50, serta diuji menggunakan confusion matrix untuk mengukur akurasi deteksi. Varian model YOLOv8s menghasilkan nilai precision sebesar 0,96, recall sebesar 0,99, F1-Score sebesar 0,97, dan nilai mAP50 sebesar 0,89. Dalam pengujian aktual, model YOLOv8s mencapai tingkat akurasi deteksi sebesar 96%, dengan tingkat kesalahan false positive dan false negative masing-masing sebesar 4%. Sistem sortir mampu melakukan proses deteksi hingga aksi sortir dalam waktu respon rata-rata kurang dari dua puluh detik. Sistem sortir otomatis berbasis CNN dengan model YOLOv8s mampu mengatasi permasalahan dengan parameter akurasi sebesar 96% dan waktu respon rata-rata kurang dari satu detik, sehingga memungkinkan implementasi pada skala industri. Model YOLOv5l dan YOLOv5m dapat menjadi alternatif apabila prioritas utama adalah kecepatan inferensi yang lebih tinggi dengan sedikit toleransi terhadap penurunan akurasi.
======================================================================================================================================
The manual sorting of bent sausage products in the food industry often faces challenges due to the subjectivity of the operator's assessment as well as fatigue factors, which lead to variations in product quality. This research designs an automatic sorting system based on Convolutional Neural Network (CNN) using the YOLOv5 (s, m, l, x) and YOLOv8 (s, m, l, x) models to detect and classify normal sausages and bent sausages. The research dataset consists of 1,461 sausage images collected from a prototype conveyor and a production conveyor, then going through annotation and augmentation stages to enhance image variability in model training. The model training was conducted in Google Colaboratory with evaluation parameters including precision, recall, F1-Score, and mAP50, and was tested using a confusion matrix to measure detection accuracy. The YOLOv8s model variant produced a precision value of 0.96, recall of 0.99, F1-Score of 0.97, and mAP50 value of 0.89. Image data. In actual testing, the YOLOv8s model achieved a detection accuracy rate of 96%, with false positive and false negative error rates of 4% each. The sorting system is capable of performing the detection process up to the sorting action with an average response time of less than twenty seconds. The CNN based automatic sorting system with the YOLOv8s model can address issues with an accuracy parameter of 96% and an average response time of less than one second, thus allowing for implementation on an industrial scale. The YOLOv5l and YOLOv5m models can serve as alternatives when the main priority is higher inference speed with some tolerance for reduced accuracy.
Item Type: | Thesis (Diploma) |
---|---|
Uncontrolled Keywords: | CNN, YOLOv5, YOLOv8, sosis bengkok, deteksi objek, sistem sortir, pengolahan citra digital, CNN, YOLOv5, YOLOv8, bent sausage, object detection, sorting system, digital image processing |
Subjects: | T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms |
Divisions: | Faculty of Vocational > 36304-Automation Electronic Engineering |
Depositing User: | Muhammad Andi Kurniawan |
Date Deposited: | 08 Aug 2025 02:56 |
Last Modified: | 08 Aug 2025 02:56 |
URI: | http://repository.its.ac.id/id/eprint/127986 |
Actions (login required)
![]() |
View Item |