Mobile Application for Soil Classification and Plant Recommendation Using Transfer Learning

Agriculture plays a crucial role in the Philippines, yet paradoxically, farmers and fisherfolk remain among the nation's poorest populations. This disparity stems from two significant challenges: limited financial resources to acquire essential inputs, such as fertilizers, and the lack of access to modern technologies that enhance productivity and efficiency. Consequently, aging farmers are increasingly reluctant to encourage their children to pursue farming, posing a significant threat to national food security. Soil health is fundamental to thriving agriculture, yet traditional soil testing methods are time-consuming and costly, often leading to inefficient trial-and-error approaches. This study proposes a mobile application designed to provide affordable, accessible, and accurate soil type predictions, thereby empowering users with essential agricultural knowledge. The dataset, comprising six soil types—chalky, clay, loamy, peaty, sandy, and silty—was collected from various municipalities in Biliran. The soil classification model was developed using a Convolutional Neural Network (CNN), achieving 95.28% accuracy during training and demonstrating 83.30% accuracy when tested with real soil images. This innovation has the potential to attract a new generation of farmers by making farming more informed and accessible.