Integrasi Dominating Set dalam Penerapan YOLO untuk Identifikasi Penyakit VSD pada Tanaman Kakao

Abstract

This research is an application of machine learning in the field of precision agriculture. It aims to obtain an object detection model to detect VSD disease in cocoa plants. The dataset used is primary data, which consists of 2 classes, namely the healthy class and the VSD class, with 1250 images in each class. The method used is the YOLOv1-YOLOv8 network model. Each YOLO network is applied in the training data process with 4000 iterations. The mean average precision (mAP) value and the number of iterations where the early stopping point (ESP) occurs in each YOLO network model are used as model evaluations. The best results were obtained from the YOLOv5 and YOLOv8 network models. YOLOv5 has the highest mAP value of 99.006% with an early stop point (ESP) in the 75th iteration, while YOLOv8 has the highest mAP value of 98.929% with ESP in the 20th iteration. The results of the model analysis show that YOLOv8 is superior to YOLOv5 because in the 20th iteration, YOLOv8 has achieved ESP, while YOLOv5 has achieved ESP in the 75th iteration, with a MAP value that is not much different. The next stage is to identify the level of spread of VSD disease in cocoa plants in a field. The graph dominating set theory is used. Field is represented as a king graph 𝑃𝑚 ⊠ 𝑃𝑛. Based on the results of the observations, the severity level on land 1 was low, land 2 and land 3 was high so the treatment must be carried out immediately on land 2 and 3.