Optimasi hyperparameter yolo26 dalam mendeteksi tingkat kematangan buah kelapa sawit
Penerbit : FTI - Usakti
Kota Terbit : Jakarta
Tahun Terbit : 2026
Pembimbing 1 : Abdul Rochman
Kata Kunci : Keywords: YOLO26; object detection; computer vision; oil palm; deep learning
Status Posting : Published
Status : Lengkap
| No. | Nama File | Hal. | Link |
|---|---|---|---|
| 1. | 2026_SK_STF_064002200029_Halaman-Judul.pdf | ||
| 2. | 2026_SK_STF_064002200029_Surat-Pernyataan-Revisi-Terakhir.pdf | ||
| 3. | 2026_SK_STF_064002200029_Surat-Hasil-Similaritas.pdf | ||
| 4. | 2026_SK_STF_064002200029_Halaman-Pernyataan-Persetujuan-Publikasi-Tugas-Akhir-untuk-Kepentingan-Akademis.pdf | ||
| 5. | 2026_SK_STF_064002200029_Lembar-Pengesahan.pdf | ||
| 6. | 2026_SK_STF_064002200029_Pernyataan-Orisinalitas.pdf | ||
| 7. | 2026_SK_STF_064002200029_Formulir-Persetujuan-Publikasi-Karya-Ilmiah.pdf | ||
| 8. | 2026_SK_STF_064002200029_Bab-1-Pendahuluan.pdf | ||
| 9. | 2026_SK_STF_064002200029_Bab-2-Landasan-Teori.pdf |
|
|
| 10. | 2026_SK_STF_064002200029_Bab-3-Metodologi-Penelitian.pdf |
|
|
| 11. | 2026_SK_STF_064002200029_Bab-4-Analisis-dan-Pembahasan.pdf |
|
|
| 12. | 2026_SK_STF_064002200029_Bab-5-Kesimpulan-dan-Saran.pdf | ||
| 13. | 2026_SK_STF_064002200029_Daftar-Pustaka.pdf | ||
| 14. | 2026_SK_STF_064002200029_Lampiran.pdf | 1 |
|
P Proses penentuan buah sawit saat ini kebanyakan masih dilakukan secara manual, yang memiliki keterbatasan dari segi waktu, efisiensi, dan konsistensi. penelitian ini bertujuan untuk menentukan konfigurasi hyperparameter yang mampu menghasilkan performa yang baik pada model yolo26 dalam mendeteksi tingkat kematangan buah kelapa sawit. dataset yang digunakan adalah citra buah kelapa sawit yang sudah melalui tahap pelabelan, preprocessing dan augmentasi. berdasarkan perbandingan performa antarvarian yolo26, varian yolo26m dipilih sebagai model dasar. selanjutnya dilakukan optimasi hyperparameter menggunakan metode optimasi random search, grid search dan bayesian optimization dengan kombinasi weight decay, learning rate, batch size, dan optimizer. berdasarkan matriks evaluasi dan karakteristik kurva pelatihan, serta mempertimbangkan biaya dan waktu komputasi, metode bayesian optimization dipilih sebagai metode optimasi yang digunakan untuk memperoleh konfigurasi hyperparameter terbaik dengan nilai accuracy sebesar 0.9731, precision sebesar 0,993, recall sebesar 0,994, map50 sebesar 0,994, dan map50-95 sebesar 0,907. penelitian ini diharapkan dapat menjadi referensi dalam penerapan optimasi hyperparameter pada model yolo26 serta mendukung pengembangan sistem deteksi otomatis tingkat kematangan buah kelapa sawit berbasis computer vision.
D Determining the oil palm fresh fruit bunches (ffbs) ripeness level is still done mostly manually, and because of that, there are limits in terms of time, efficiency, and overall consistency. in this study, we want to find the best hyperparameter setup to boost the detection performance of the yolo26 model for oil palm fruit ripeness. the dataset uses oil palm fruit images, then they went through labeling, preprocessed, and augmented. after that, we compared the performance of several yolo26 versions, and yolo26m was picked as the main or base model. next, the hyperparameter optimization process was carried out with grid search, random search, and bayesian optimization, while tuning the learning rate, batch size, weight decay, and optimizer. looking at the evaluation metrics, the training loss curves, plus thinking about computational cost and training time, bayesian optimization ended up being the more suitable choice. the optimized model shows a accuracy of 0.9731, precision of 0.993, a recall of 0.994, an map50 of 0.994, and an map50–95 of 0.907. overall, the findings here give some useful views about how hyperparameter tuning affects the yolo26 model performance, and they could be used as a sort of reference for building computer-vision-based automatic systems to detect oil palm fruit ripeness.