| dc.contributor.advisor | Julianto, Mochamad Tito | |
| dc.contributor.advisor | Khatizah, Elis | |
| dc.contributor.author | Zahrani, Rahma Alya | |
| dc.date.accessioned | 2026-07-21T14:57:02Z | |
| dc.date.available | 2026-07-21T14:57:02Z | |
| dc.date.issued | 2026 | |
| dc.identifier.uri | http://repository.ipb.ac.id/handle/123456789/175295 | |
| dc.description.abstract | Penyakit tanaman padi seperti bacterial blight, blast, brown spot, dan tungro kerap dijumpai di lapangan dan dapat menurunkan kuantitas maupun kualitas hasil panen sehingga identifikasi terhadap keempatnya menjadi penting untuk membatasi kerugian yang ditimbulkan. Dalam upaya identifikasi keempat penyakit tanaman padi tersebut dapat dilakukan dengan klasifikasi citra menggunakan model ResNet-50 dengan pendekatan transfer learning. Salah satu kendala selama proses pelatihan adalah distribusi kelas yang tidak seimbang. Untuk mengatasi hal ini, diterapkan
class weight pada fungsi loss agar kontribusi setiap kelas terhadap proses pelatihan menadi lebih proporsional. Penelitian ini bertujuan menganalisis pengaruh penerapan class weight dan penggunaan algoritma optimasi Adam dan Stochastic
Gradient Descent with Momentum (SGD-M) terhadap kinerja model. Hasil menunjukkan bahwa skenario terbaik adalah optimizer Adam dengan class weight yang menghasilkan accuracy 98,99% dan f1-score 99,01%, dibandingkan optimizer
Adam tanpa class weight sebesar 98,85% dan optimizer SGD-M dengan class weight sebesar 98,60%. | |
| dc.description.abstract | Rice plant diseases such as bacterial blight, blast, brown spot, and tungro are commonly found in the field and can reduce both the quantity and quality of harvest yields, making identification of these four diseases important to limit the
resulting losses. Identification of these four rice plant diseases can be carried out through image classification using a ResNet-50 model with a transfer learning approach. One challenge encountered during the training process is class imbalance in the data distribution. To address this, class weight is applied to the loss function so that each class contributes more proportionally during training. This study aims to analyze the effect of applying class weight and using the Adam and Stochastic Gradient Descent with Momentum (SGD-M) optimization algorithms on model performance. The results show that the best-performing scenario is the Adam optimizer with class weight, achieving an accuracy of 98.99% and an F1-score of 99.01%, compared to Adam without class weight at 98.85% and SGD-M with class weight at 98.60%. | |
| dc.description.sponsorship | | |
| dc.language.iso | id | |
| dc.publisher | IPB University | id |
| dc.title | Penerapan Class Weight pada Model ResNet-50 untuk Klasifikasi Citra Penyakit Tanaman Padi | id |
| dc.title.alternative | Application of Class Weight on ResNet-50 Model for Rice Plant Disease Image Classification | |
| dc.type | Skripsi | |
| dc.subject.keyword | Class weight | id |
| dc.subject.keyword | image classification | id |
| dc.subject.keyword | ResNet50 | id |
| dc.subject.keyword | rice plant disease | id |
| dc.subject.keyword | transfer learning | id |
| dc.subtype | Undergraduate Theses | |