DEVELOPINGBACTERIAL WILT CLASSIFICATION MODELON ENSET CROP USING DEEP LEARNING.

Abstract

Ethiopia has significant agricultural potential, particularly in the cultivation of Enset, a monocarp perennial crop belonging to the family of Musaceae, which is necessary for food security in Ethiopia, especially in the southern part, serving as a basic food for approximately 20 million individuals. Nevertheless, its production is susceptible to a number of diseases brought on by bacteria, fungus, and viruses, making it more difficult to classify and treat disorders. Given the large size of Enset crops, it is often ineffective for plant pathologists to examine each plant individually. Previous studies mostly concentrated on identifying tasks without utilizing cutting-edge technologies for efficient disease control. To solve the problem this study proposes the development of a deep learning model to automatically classify Enset diseases, specifically bacterial wilt. With the assistance of agricultural specialists from different farms, the researcher gathered a dataset of 2,000 images of Enset plants that were both healthy and bacterial wilted. This study used a mixed research approach using an experimental research design. Python on Google Colab was used for data analysis, prototype creation, and model building in this study. The dataset was split into 70% for training, 20% for validation, and 10% for testing. The strategy made use of convolutional neural network (CNN) and evaluated its performance against pre-trained models like ResNet50 and VGG19. During training, we used data augmentation approaches to improve the model's performance. The results showed that, in spite of difficult circumstances such as shifting lighting, complicated backgrounds, and various orientations, our model was able to obtain training accuracy of 99.94% and a test accuracy of 97% of the VGG19 model. So VGG19 classifier performs better than the other classifiers, according to the experiment's final results. This outcome ultimately leads to the development and testing of a system prototype that can categorize disease of Enset leafs.