Ade Arya Anggara, Abdurrahman Ridho, Cut Mutia

Analisa Penyakit Pada Tanaman Cabai Merah (Capsicum annuum L) Dengan Membandingkan Tingkat Akurasi Menggunakan Metode Convolutional Neural Network (CNN) dan K-Nearest Neighbor (KNN)

Introduction

Analisa penyakit pada tanaman cabai merah (capsicum annuum l) dengan membandingkan tingkat akurasi menggunakan metode convolutional neural network (cnn) dan k-nearest neighbor (knn). Analisis penyakit cabai merah menggunakan CNN (VGG16) & KNN. Bandingkan akurasi identifikasi penyakit tanaman. CNN mencapai 86%, lebih akurat dari KNN. Solusi computer vision.

Abstract

Cabai merah (Capsicum annuum L) adalah bumbu pokok yangada di hampir setiap masakan Indonesia dan memiliki nilai jualtinggi dibandingkan tanaman lain. Namun, banyak faktorserangan yang terjadi baik biotik maupun abiotik, sehinggadapat menyebabkan gagalnya panen. identifikasi yang cepatdapat mengurangi risiko kegagalan panen ini. Salah satusolusinya adalah menggunakan teknologi computer visionuntuk mengklasifikasikan citra penyakit tanaman cabai merah.Dalam penelitian ini, digunakan metode Convolutional NeuralJaringan (CNN) dan K-Nearest Neighbor (KNN) untukmembuat model klasifikasi yang dapat memprediksi penyakitpada tanaman cabai merah. Arsitektur VGG16 digunakandalam model CNN dan KNN untuk dapat membantu dalamekstraksi fitur dan pengenalan pola pada citra. Hasilnya, modelCNN memberikan akurasi sebesar 85.46% (dibulatkanmenjadi 86%), sementara model KNN menghasilkan akurasisebesar 81%. Model evaluasi dilakukan dengan menggunakanmetode kerumitan matriks. Metode kekacauan matriks ini akanmenghitung nilai akurasi, presisi, recall dan F1-score darimodel masing-masing. Berdasarkan perhitungan yang telahdilakukan, diperoleh nilai akurasi tertinggi berada pada modelCNN.

Review

This paper addresses a highly relevant agricultural challenge: the rapid and accurate identification of diseases in red chili plants (Capsicum annuum L.), a staple crop in Indonesia. The authors propose a computer vision-based solution, comparing the performance of Convolutional Neural Networks (CNN) and K-Nearest Neighbor (KNN) methods for classifying plant diseases. The methodology employs the VGG16 architecture, which is a strong choice for feature extraction and pattern recognition in images, utilized for both the CNN model and in extracting features for the KNN model. The clear objective of comparing these two established techniques to achieve early disease detection is commendable, setting a solid foundation for the research. The study presents a clear comparison of the two classification approaches. The CNN model, leveraging the VGG16 architecture, achieved a notable accuracy of 85.46% (rounded to 86%), outperforming the KNN model which yielded an accuracy of 81%. This demonstrates the superior capability of deep learning for complex image classification tasks in this domain. The evaluation using a confusion matrix to derive metrics such as accuracy, precision, recall, and F1-score indicates a robust assessment of model performance. However, a significant omission in the abstract is the lack of detail regarding the dataset used, specifically the number of images and the types of diseases classified. This information is crucial for understanding the scope, generalizability, and practical applicability of the developed models. In conclusion, this research successfully highlights the potential of CNNs for automated disease diagnosis in red chili plants, offering a promising tool to reduce crop failure and enhance agricultural productivity. The finding that CNN significantly outperforms KNN, especially when both leverage a powerful feature extractor like VGG16, is a valuable contribution to the field. For future work, it would be beneficial to expand upon the dataset size and diversity, including more specific disease types and varying environmental conditions, to further validate and improve model robustness. Exploring real-time implementation strategies and comparing the proposed models with more recent deep learning architectures could also pave the way for practical deployment in agricultural settings.

Full Text

Comments

You need to be logged in to post a comment.