Detection of Disease in Platycerium Ornamental Plant Leaves Using Yolo 12

Abstract

Platycerium is an epiphytic ornamental plant with high aesthetic and economic value, thus requiring proper care. Identifying Platycerium leaf diseases based on visual symptoms often requires precision and experience, thus necessitating an image-based automated approach. This study aims to develop a Platycerium leaf disease detection model using the deep learning-based YOLO method. The model was developed using Kaggle Notebook with P100 GPU support. The dataset used consisted of three disease classes, namely Bacterial Leaf Spot, Fern Scale, and Rizoctonia Blight. Model training was carried out with variations in the number of epochs of 50, 75, and 100 epochs, and evaluated using the Precision, Recall, and Mean Average Precision (mAP) metrics. The results showed that training with 50 epochs gave the best results with a mAP50 value of 0.953 and mAP50–95 of 0.577. Testing using test data and data outside the dataset showed that the model was able to detect Platycerium leaf disease in test images by displaying bounding boxes and class labels. Based on these results, the YOLO model developed can be used as an image-based approach for detecting Platycerium leaf disease and can be further developed.