Real-time detection of fruit ripeness using the YOLOv4 algorithm
<p align="justify"><span lang="IN">Indonesia is a country that has many types of commodities of various kinds of fruit. Based on BPS data, fruit production reached 22 million tons in 2019, an increase of 5% compared to 2018. However, with such a large production volum...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Universitas Sultan Ageng Tirtayasa
2021-11-01
|
| Series: | Teknika |
| Subjects: | |
| Online Access: | https://jurnal.untirta.ac.id/index.php/ju-tek/article/view/12254 |
| Summary: | <p align="justify"><span lang="IN">Indonesia is a country that has many types of commodities of various kinds of fruit. Based on BPS data, fruit production reached 22 million tons in 2019, an increase of 5% compared to 2018. However, with such a large production volume, most fruit inspection (sorting) processes in Indonesia are still carried out manually by humans. The manual process requires a lot of time and energy resources, is prone to inconsistencies and inaccuracies. With the help of computer vision, the automation process is expected to be able to eliminate manual processes to reduce costs and increase efficiency and accuracy. In this study, the method used is the YOLOv4 algorithm. The algorithm is applied to detect banana ripeness automatically. The algorithm training process is carried out using 369 banana images divided into two classes, and the testing process is carried out on videos captured in real-time. Based on the research results, the best average accuracy rate is 87.6%, and the video processing speed is 5 FPS (frames per second) using a single-GPU architecture.</span></p><p align="justify"> </p><p align="justify">Indonesia merupakan negara yang memiliki banyak jenis komoditi berbagai macam buah-buahan. Berdasarkan data BPS, produksi buah mencapai 22 juta ton pada 2019 atau meningkat 5% dibandingkan 2018. Namun dengan volume produksi sebesar itu, sebagian besar proses pemeriksaan (<em>sorting</em>) buah di Indonesia masih dilakukan secara manual oleh manusia. Proses manual membutuhkan banyak waktu dan sumber daya, rentan terhadap inkonsistensi dan ketidakakuratan. Proses otomatisasi dengan bantuan <em>computer vision</em> diharapkan dapat menghilangkan proses manual sehingga dapat menekan biaya, meningkatkan efisiensi dan akurasi. Dalam penelitian ini, metode yang digunakan adalah algoritma YOLOv4. Algoritma ini diterapkan untuk mendeteksi kematangan pisang secara otomatis. Proses pelatihan algoritma dilakukan dengan menggunakan 369 citra pisang yang dibagi menjadi dua kelas dan proses pengujian dilakukan pada video yang ditangkap secara real time. Berdasarkan hasil penelitian, tingkat akurasi rata-rata terbaik adalah 87,6% dan kecepatan pemrosesan video adalah 5 FPS (<em>frame per seconds</em>) dengan menggunakan arsitektur <em>single-</em>GPU.</p> |
|---|---|
| ISSN: | 1693-024X 2654-4113 |