IMPLEMENTATION OF PLANT DISEASE CLASSIFICATION MODEL USING TRANSFER LEARNING AND DCGAN DATA AUGMENTATION IN ANDROID BASED APPLICATION (CASE STUDY ON TOMATO LEAVES)

• Main Author: Justine William, Christopher
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/73915
• Institution: Institut Teknologi Bandung

Early detection of plant diseases is crucial for preventing further spread and damage. With advancements in computer vision technology, the deep learning approach has emerged as an effective method for disease detection in plants. However, to make these models accessible to a wider audience, it is important to develop models that can run easily on simpler devices like mobile gadgets. The objective of this Final Project is to develop a plant disease classifier model using the transfer learning method and the DCGAN data augmentation technique in an android-based application. For this Final Project, a case study focusing on tomato leaf disease from the PlantVillage dataset is utilized. The transfer learning approach, which leverages pre-trained models, is chosen over developing models from scratch due to its ability to accelerate the training process by utilizing existing knowledge. Furthermore, the DCGAN data augmentation technique helps address data limitations by generating additional data, thereby enhancing dataset balance and variation. Based on the test and evaluation results, the model achieved the highest accuracy of 97.83% when trained and tested on the PlantVillage dataset. This model with the highest accuracy is then implemented in an android application for classifying diseases in tomato plant leaves.