Mobilenet-based Transfer Learning for Detection of Eucalyptus Pellita Diseases

Currently, the pulp industry in Indonesia is ranked eighth world and paper sixth world. One of advantages supporting that has a large Industrial Plantation Forest (HTI) where plants for raw materials originate. Eucalyptus pellita species belonging to Myrtaceae family one priority Forests because its adaptability wood can be used as material pulp. this type found mainly Kalimantan Sumatra. This shows good growth stem shape, speed quality high germination shorter cutting cycle about 7-8 years so it quickly harvested. Prevention treatment leaf disease main processes planting. Early diagnosis accurate recognition Pellita control spread reduce production costs costs. Disease detection on leaves done automatically faster by utilizing digital image processing artificial intelligence. In study, we propose method with Deep Learning architecture. Our proposed based pre-trained transfer learning using MobileNet. Image datasets from PT. Surya Hutani Jaya's land East were train model. The dataset divided into three classes 1 class healthy 2 are sick leaves, namely Xanthomonas Bacteria Cylindrocladium Fungi. With ratio 70: 20: 10 number training 2370, validation 591, Testing 177. Hyperparameter scenarios carried out MobileNet model optimize performance dataset. experimental results show fairly accuracy, reaching 98%.