Klasifikasi Hama dari Citra Daun Brokoli Menggunakan Convolutional Neural Network Efficientnet-B0

Abstract

Broccoli (Brassica oleracea L.) is a vegetable that belongs to the cabbage family. This vegetable is well known for its abundance of nutrients and nutrients, including vitamin C, vitamin K, iron, and antioxidant compounds that are useful for health. Every year many broccoli crops fail due to pests and the main way to deal with them is by spraying pesticides. In controlling pests, most farmers spray pesticides without paying attention to the right dose, time, method, and target. This action causes negative impacts such as killing organisms that are not the target of pests. To overcome these problems, this research utilizes the CNN method to classify pests that attack broccoli plants. The dataset utilized in this research is in the form of images that have been taken directly from broccoli farms in the Berastagi area. The test results show that the CNN model produces the highest level of accuracy, namely training accuracy of 95.69% with a training loss of 0.12 and validation accuracy of 98.96% with a validation loss of 0.06 with a total of 75 epochs in model training. In evaluating the model using confusion matrix, the accuracy value is 96.56%, precision value is 96.55%, recall value is 96.61%, and 96.57% F1-score in pest classification. This research contributes to the development of technology to support broccoli farmers in overcoming the problem of pest attacks that can threaten crop yields and to facilitate users, this system is designed for Android devices directly.