E'MELON: Android-Based Watermelon Leaf Disease Detection and Remedy Using TensorFlow and CNN

Abstract

The early detection of watermelon leaf diseases is crucial for effective crop management, yet traditional methods are labor-intensive and require technical expertise. This study aimed to develop an Androidbased application using TensorFlow and Convolutional Neural Networks (CNN) to detect and remedy watermelon leaf diseases. The CNN-based software leverages image classification techniques to enhance accuracy and reduce the complexity of previous algorithms. The application was evaluated by 19 randomly selected participants using the ISO/IEC 9126 Software Quality Model, resulting in a weighted mean accuracy score of 4.42, which corresponds to good, which highlights that most respondents agree that the Android application effectively identifies watermelon diseases and provides appropriate treatment recommendations. This rating was determined based on specific criteria such as functionality and reliability. However, the classifier's performance needs improvement, particularly in distinguishing between similar disease symptoms. Feedback from watermelon experts and farmers indicated that while the application is promising, enhancements needed on the delivery of specific data for certain disease categories are needed for ease of understanding. The findings highlight the potential of mobile-based applications in agricultural disease management and the need for further refinement to achieve higher precision.