Enhancing Agricultural Precision: A Hybrid CNN-SVM Approach for Multidimensional Weed Disease Classification

A new category of weed disease classification in precision agriculture based on a hybrid model using convolutional neural networks (CNN) and support vector machine (SVM) has been presented in the proposed study. In line with overcoming the shortcomings of conventional disease detection approaches, this study introduces a machine learning solution that improves accuracy and productivity in the identification of weed diseases. The hybrid model utilizes the strengths of CNNs for feature extraction and its power with SVM in classification, yielding a stable system that enables multilevel classification of weed disease severity. Creation of a complete collection of weed disease image datasets for the training, validation, and testing phase. The findings from the study show that the use of the model outperforms standalone CNN & SVM models as it achieves a substantial increase in 96.21% accuracy across all the severity levels. The model also uses Grad-CAM visualizations to ensure interpretability and enable a better understanding of the decision-making process. The results imply that the proposed hybrid CNN-SVM model is both efficient in terms of weed disease classification and feasible for use across a range of agricultural settings. This research provides a valuable tool for the development of precision agriculture, which enables early and precise intervention in plant diseases that positively impact sustainable farming. Future work may include increasing the application potential of the model and investigating its integration into automated agricultural systems.