Rice Pest and Disease Classification using Transfer Learning Inception V3 with data augmentation.

Rice food security is crucial amidst rapid population growth and global climate instability. Rice is one of the primary food sources for most of the world’s population, especially in the Asian region. It is necessary to carry out food security, especially by helping farmers identify pests and diseases. Computers can help identify conditions, one of which is through pattern recognition of pests and diseases. The method that can be used is the Deep learning convolutional neural network (CNN). This research used twelve types of pests and diseases that often attack rice plants: mealybugs, leafhoppers, weeds, stem borers, rats, and birds. The conditions that often attack are leaf blight, brown spot, blast, striped leaves, tungro, and healthy. This research uses data taken from agricultural locations and secondary data. Each class consists of 50-70 images. The proposed method is Inception transfer learning with augmentation. The contribution offered is an architecture that does not require training computation time and has high accuracy. The test scenario results show that the average model accuracy is 97.75, with a computational training time of around 45 minutes. MSE 0.0018, RMSE 0.0426 dan MAE 0.0018. Testing using experimental data takes 1-2 seconds, with confidence accuracy ranging from 95.8%-99.2%.