Advanced Techniques for Sweet Potato Leaf Disease Detection: A CNN-SVM Hybrid Approach

This study provides a thorough analysis of a system of classification intended to identify five different illnesses that impact sweet potato leaves. The model demonstrates Notable performance metrics, which uses a complex architecture with three convolution layers, three max-pooling layers, and one fully linked layer. The accuracy of optimistic predictions is indicated by precision values between 88.73% and 91.78%, while the proportion of real positives properly anticipated is indicated by recall values between 86.11% and 91.78%. The F1-Score falls between 87.32% and 91.78%, weighing recall and precision. The model's accuracy, which ranges from 95% to 97% for each illness class, is impressive. The model's overall efficacy is further supported by aggregated measures, where macro, weighted, and micro averages consistently show excellent ratings. In order to improve agricultural productivity and food security, this study highlights the potential of using cutting-edge convolutional neural network architectures for the accurate and dependable classification of sweet potato leaf diseases.