Please use this identifier to cite or link to this item:
https://repositori.uma.ac.id/handle/123456789/26468
Title: | Analisis Kinerja Hyperparameter Convolutional Neural Network pada Kasus Klasifikasi Penyakit Daun Cabai |
Other Titles: | Performance Analysis of Convolutional Neural Network Hyperparameters in the Case of Chili Leaf Disease Classification |
Authors: | Setyadi, Rahmat Arief |
metadata.dc.contributor.advisor: | Rahman, Sayuti |
Keywords: | hyperparameter tuning;cnn;cabai;deep learning;chili |
Issue Date: | 26-Aug-2024 |
Publisher: | UNIVERSITAS MEDAN AREA |
Series/Report no.: | NPM;208160026 |
Abstract: | Penyakit yang menyerang daun pada tanaman cabai merupakan tantangan dalam pertanian modern yang dapat diatasi dengan teknologi pengolahan citra dan kecerdasan buatan. Penelitian ini mengusulkan metode hyperparameter tuning, citra daun cabai digunakan dalam dua kondisi: tanpa augmentasi dan dengan augmentasi. Pada arsitektur ResNet101 tanpa augmentasi dataset, akurasi pelatihan mencapai 98.3529% dan akurasi validasi 89.7196%. Dengan augmentasi, akurasi pelatihan sebesar 97.8261% dan akurasi validasi meningkat signifikan menjadi 97.1831% Perbandingan kinerja arsitektur CNN lainnya menunjukkan ResNet50 dengan akurasi pelatihan 98.4724% dan validasi 97.4178%, sementara ResNet101 dengan hyperparameter tuning mencapai akurasi pelatihan 97.6498% dan validasi tertinggi 98.1221%. hyperparameter dengan fungsi aktivasi Tanh dengan learning rate schedule ReduceLR pada ResNet101 menghasilkan akurasi pelatihan 97.7673% dan validasi tertinggi 99.5305% Hasil evaluasi menunjukkan bahwa hyperparameter tuning meningkatkan akurasi klasifikasi secara signifikan. Metode ini berkontribusi dalam pengembangan sistem deteksi dini penyakit pada tanaman cabai secara efektif dan efisien. Hyperparameter tuning juga menunjukkan potensi yang besar untuk diterapkan pada berbagai jenis tanaman lain dalam pertanian membantu petani dalam mengidentifikasi dan mengelola penyakit tanaman dengan lebih baik. Diseases that attack the leaves of chili plants are a challenge in modern agriculture that can be overcome with image processing technology and artificial intelligence. This research proposes a hyperparameter tuning method, chili leaf images are used in two conditions: without augmentation and with augmentation. On the ResNet101 architecture without dataset augmentation, training accuracy reached 98.3529% and validation accuracy 89.7196%. With augmentation, the training accuracy is 97.8261% and the validation accuracy increases significantly to 97.1831%. Comparison of the performance of other CNN architectures shows ResNet50 with a training accuracy of 98.4724% and validation of 97.4178%, while ResNet101 with hyperparameter tuning achieves a training accuracy of 97.6498% and the highest validation of 98.1221%. hyperparameter with Tanh activation function with ReduceLR learning rate schedule on ResNet101 produces training accuracy of 97.7673% and highest validation of 99.5305%. Evaluation results show that hyperparameter tuning increases classification accuracy significantly. This method contributes to the development of an effective and efficient early detection system for diseases in chili plants. Hyperparameter tuning also shows great potential to be applied to various other types of crops in agriculture helping farmers in better identifying and managing plant diseases. |
Description: | 68 Halaman |
URI: | https://repositori.uma.ac.id/handle/123456789/26468 |
Appears in Collections: | SP - Informatic Engineering |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
208160026 - Rahmat Arief Setyadi - Fulltext.pdf | Cover, Abstract, Chapter I, II, III, V, Bibliography | 1.44 MB | Adobe PDF | View/Open |
208160026 - Rahmat Arief Setyadi - Chapter IV.pdf Restricted Access | Chapter IV | 361.67 kB | Adobe PDF | View/Open Request a copy |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.