Abstract

This research develops an automated classification system for purple sweet potato leaf conditions using Convolutional Neural Network (CNN). The dataset comprises 1773 images spanning three leaf conditions: yellow (473 images), brown (535 images), and healthy (765 images). Data was split into 80% training and 20% validation, with input images sized at 150x150 pixels processed through augmentation techniques. The developed CNN model achieved 100% accuracy on validation data with minimal error rate. The confusion matrix demonstrated the model's success in classifying 76 yellow leaf samples, 103 brown leaf samples, and 176 healthy leaf samples without misclassification. This research proves CNN's effectiveness in classifying purple sweet potato leaf conditions, enabling early disease detection for more appropriate treatment.