<p>Plant disease identification traditionally relies on manual visual inspection, a process that is labor-intensive, error-prone, and dependent on specialized expertise. In precision agriculture, early and accurate disease detection is essential for timely intervention and sustainable crop management. Existing frameworks largely focus on disease classification and provide limited insight into symptom severity, often functioning as black-box models. This study introduces an interpretable, hierarchical three-module framework for strawberry leaf scorch analysis that integrates deep learning features, human-interpretable symptom diagnosis, and severity prediction within a unified system-level design. In Module-1 (Disease Detection), a fine-tuned deep learning model performs binary classification of leaf images into healthy and leaf scorch-affected classes, achieving 100% accuracy on the standardized test set under controlled experimental conditions. Only infected leaves proceed to Module-2 (Symptom Diagnosis), which assesses five interpretable symptom attributes: leaf greenness, infection level, spot density, spot distribution, and texture. Each symptom is assigned a severity stage using a predefined rubric validated by plant pathology experts. Module-3 (Severity Prediction) integrates these outputs through weighted majority voting, achieving 97.14% severity prediction accuracy against the internally defined rubric. Future work is directed toward adaptation and rigorous validation in diverse field environments.</p>