Identifikasi Tingkat Kematangan Buah Kelapa Sawit Menggunakan Kombinasi Model You Only Look Once Version 8 dan Convolutional Neural Netwrok

Azahari, Muchammad Aquila (2024) Identifikasi Tingkat Kematangan Buah Kelapa Sawit Menggunakan Kombinasi Model You Only Look Once Version 8 dan Convolutional Neural Netwrok. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002201153-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Industri kelapa sawit merupakan komoditas perkebunan yang mempunyai peran krusial dalam perekonomian Indonesia. Hal tersebut didukung atas capaian Indonesia sebagai negara penghasil dan eksportir kelapa sawit terbesar di dunia pada tahun 2022. Meskipun demikian, industri kelapa sawit tidak terlepas dari tantangan signifikan, khususnya dalam upaya menghasilkan minyak kelapa sawit mentah (CPO) yang berkualitas. Tingkat kematangan buah kelapa sawit menjadi faktor penting dalam mencapai kualitas CPO yang optimal. Metode mutakhir untuk mengidentifikasi tingkat kematangan buah kelapa sawit telah banyak dikembangkan, namun model tersebut belum memperoleh hasil yang memuaskan karena karakteristik kompleks dari buah kelapa sawit. Model tersebut masih dihadapkan dengan sejumlah tantangan kritis seperti pengurangan gangguan, ekstraksi fitur yang relevan, dan pengecualian fitur yang berlebihan. Pada penelitian ini, digunakan kombinasi model You Only Look Once version 8 dan Convolutional Neural Network untuk mengidentifikasi tingkat kematangan buah kelapa sawit. Terdapat dua tahapan utama dalam penelitian ini. Tahap pertama yakni pelatihan model dengan menggunakan 2 skenario pelatihan, dan tahapan kedua adalah pengujian model terbaik menggunakan tiga skenario pengujian. Berdasarkan hasil pelatihan model didapat model YOLOv8 dengan backbone Darknet53 memberikan peforma terbaik dengan nilai mAP sebesar 91,43% pada IoU threshold 0.5 dan nilai F-1 Score sebesar 99,08%. Model tersebut juga unggul dalam efisiensi waktu komputasi, waktu komputasi yang dibutuhkan untuk mengoptimalkan pelatihan yaitu 2,995 jam dengan 163 epoch.
========================================================================================================================
The palm oil industry is a plantation commodity that plays a crucial role in the Indonesian economy. This is supported by Indonesia's achievement as the largest palm oil producing and exporting country in the world in 2022. However, the palm oil industry is not free from significant challenges, especially in efforts to produce quality crude palm oil (CPO). The ripeness level of oil palm fruit is an important factor in achieving optimal CPO quality. Exchange methods to identify the ripeness level of oil palm fruit have been widely developed, but the model has not obtained satisfactory results due to the complex characteristics of oil palm fruit. The model still faces a number of critical challenges such as reducing interference, extracting relevant features, and reducing excessive features. In this study, a combination of the You Only Look Once version 8 model and the Convolutional Neural Network were used to identify the ripeness level of oil palm fruit. There are two main stages in this study. The first stage is model training using 2 training scenarios, and the second stage is testing the best model using three test scenarios. Based on the results of model training, the YOLOv8 model with the Darknet53 backbone provides the best performance with an mAP value of 91.43% at IoUthreshold 0.5 and an F-1 Score value of 99.08%. The model also excels in computational time efficiency, the computational time required to optimize training is 2.995 hours with 163 epochs.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Industri kelapa sawit, Buah kelapa sawit, Model deteksi objek, Model klasifikasi, You Only Look Once version 8, Convolutional Neural Network
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.6 Computer programming. Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Muchammad Aquila Azahari
Date Deposited:	05 Aug 2024 08:09
Last Modified:	05 Aug 2024 08:09
URI:	http://repository.its.ac.id/id/eprint/110534

Actions (login required)

View Item