KURNIAWAN, ERA DANI (2026) ANALISIS KINERJA ARSITEKTUR CONVOLUTIONAL NEURAL NETWORK EFFICIENTNETV2B0 DALAM KLASIFIKASI MULTIKELAS PENYAKIT DAUN TANAMAN STROBERI. S1 thesis, Universitas Mercu Buana Jakarta.
|
Text (HAL COVER)
COVER.pdf Download (558kB) | Preview |
|
![[img]](https://repository.mercubuana.ac.id/style/images/fileicons/text.png) |
Text (BAB I)
BAB I.pdf Restricted to Registered users only Download (186kB) |
|
|
Text (BAB II)
BAB II.pdf Restricted to Registered users only Download (385kB) |
|
|
Text (BAB III)
BAB III.pdf Restricted to Registered users only Download (142kB) |
|
|
Text (BAB IV)
BAB IV.pdf Restricted to Registered users only Download (411kB) |
|
|
Text (BAB V)
BAB V.pdf Restricted to Registered users only Download (25kB) |
|
|
Text (DAFTAR PUSTAKA)
DAFTAR PUSTAKA.pdf Restricted to Registered users only Download (95kB) |
|
|
Text (LAMPIRAN)
LAMPIRAN.pdf Restricted to Registered users only Download (245kB) |
Abstract
Leaf diseases in strawberry plants have the potential to reduce crop quality and productivity if not detected early. Advances in Deep Learning technology, particularly Convolutional Neural Networks (CNNs), have been widely utilized for plant disease classification based on digital images because they can identify visual characteristics directly from image data. Although various studies have demonstrated the strong performance of CNNs in plant disease classification, research specifically evaluating the performance of the EfficientNetV2B0 architecture in the multi-class classification of strawberry leaf diseases remains relatively limited. This study analyzes the performance of the EfficientNetV2B0 architecture in the multi-class classification of strawberry leaf diseases based on digital images, using the metrics accuracy, precision, recall, and F1-score. The study dataset consists of 2,615 images of strawberry leaves divided into three classes: “Healthy” (456 images), “Angular Leaf Spot” (1,050 images), and “Leaf Scorch” (1,109 images). The dataset was obtained from two public datasets on the Kaggle platform. The dataset then underwent image preprocessing and augmentation before being used to train the EfficientNetV2B0 model using a feature extraction-based transfer learning approach. The results of the study show that the EfficientNetV2B0 model achieved a training accuracy of 99.52% and a validation accuracy of 99.62%. During the testing phase, the model achieved an accuracy of 99%, with precision, recall, and F1-score values for each class ranging from 0.98 to 1.00. The evaluation results indicate that the majority of images were correctly classified, and only a small fraction of samples were misclassified across disease classes. These findings indicate that EfficientNetV2B0 delivers optimal performance in the multiclass classification of strawberry leaf diseases on the dataset used. This study provides empirical evidence of EfficientNetV2B0’s performance and can serve as a reference for further development and research in digital image-based plant disease identification systems. Keywords: EfficientNetV2B0, Convolutional Neural Network, multiclass classification, strawberry leaf diseases, Deep Learning Penyakit daun pada tanaman stroberi berpotensi mengurangi kualitas dan produktivitas hasil panen apabila tidak dideteksi secara dini. Perkembangan teknologi Deep Learning, khususnya Convolutional Neural Network (CNN), telah banyak dimanfaatkan untuk klasifikasi penyakit tanaman berbasis gambar digital karena dapat mengidentifikasi karakteristik visual langsung dari data citra. Meskipun berbagai kajian telah memperlihatkan kinerja CNN dengan baik dalam klasifikasi penyakit tanaman, kajian yang secara khusus mengevaluasi kinerja arsitektur EfficientNetV2B0 pada klasifikasi multikelas penyakit pada daun stroberi masih relatif terbatas. Penelitian ini menganalisis kinerja arsitektur EfficientNetV2B0 dalam klasifikasi multikelas penyakit daun stroberi berbasis citra digital berdasarkan metrik accuracy, precision, recall, dan F1-score.Objek penelitian berupa 2.615 citra daun stroberi yang terdiri atas tiga kelas, yaitu Healthy terdiri dari 456 citra, Angular Leaf Spot sebanyak 1.050 citra, dan Leaf Scorch sebanyak 1.109 citra. Dataset diperoleh dari dua dataset publik pada platform Kaggle. Dataset kemudian melalui tahap preprocessing dan augmentasi citra sebelum diterapkan untuk melatih model EfficientNetV2B0 melalui pendekatan transfer learning berbasis feature extraction.Hasil penelitian menunjukkan bahwa model EfficientNetV2B0 mencapai akurasi pelatihan sebesar 99,52% dan akurasi validasi sebesar 99,62%. Pada tahap pengujian, model memperoleh akurasi sebesar 99%, dengan nilai precision, recall, dan F1-score pada setiap kelas berkisar antara 0,98 hingga 1,00. Hasil evaluasi menunjukkan bahwa sebagian besar citra berhasil diklasifikasikan dengan benar dan hanya sebagian kecil sampel yang mengalami misklasifikasi antar kelas penyakit. Temuan ini menunjukkan bahwa EfficientNetV2B0 memiliki performa yang optimal dalam klasifikasi multikelas penyakit daun stroberi pada dataset yang digunakan. Kajian ini memberikan bukti empiris mengenai kinerja EfficientNetV2B0 dan dapat menjadi referensi bagi pengembangan dan penelitian lanjutan dalam sistem identifikasi penyakit tanaman berbasis citra digital. Kata kunci: EfficientNetV2B0, Convolutional Neural Network, klasifikasi multikelas, penyakit daun stroberi, Deep Learning
Actions (login required)
 |
View Item |
