Deteksi Penyakit Tanaman Cabai Menggunakan Algoritma YOLOv5 Dengan Variasi Pembagian Data

Abstract

Rapid technological developments have resulted in various innovative techniques that help humans, including object detection which functions to identify each element in an image. Object detection is often used to overcome problems that occur because of its ability to identify each element in the image. One of the problems that is often encountered is a decrease in agricultural income due to disease in chili plants. The maintenance of chili plants has various obstacles including the impact of weather which causes the development of diseases and pests so that chili production has decreased. By implementing the object detection, farmers can easily identify diseases that attack chili plants through pictures so that chili disease can be treated more quickly. This study uses the YOLOv5 algorithm to test the performance of the model in identifying diseases in chili plants. Pictures were taken using a cellphone camera with dimensions of 3472x3472 pixels. The amount of image data used is 430 data. Image data is divided into 3 parts, namely train data, validation data, and test data. To get the best model, this study also conducted three experiments with different distribution of data. Experiment 1 with a division of 70:20:10, experiment 2 with a division of 75:15:10, and experiment 3 with a division of 80:10:10. From the experiments carried out, the best results were obtained, namely in experiment 3 with an average value obtained in the test of 0.947 with a translation of the precision, recall, and mAP values, namely 0.946, 0.936, and 0.959 respectively.