Application of deep transfer learning for foliar disease detection in Solanaceous vegetables for sustainable production

Solanaceous vegetables are vital staple crops that contribute significantly to global food security; however, plant pathogens cause annual crop losses of 20–40%, resulting in estimated economic losses of USD 220 billion worldwide. This study proposes an enhanced Solanaceae Leaf Disease Classification Network (SLDCNet) leveraging deep transfer learning for early-stage identification of foliar diseases in Solanaceous vegetables. A comprehensive dataset comprising 5,211 high-resolution images representing nine disease classes was collected from farmers’ fields in the Kapurthala district, Punjab, India, under diverse field conditions. The dataset was pre-processed and augmented through rotation, shifting, and shearing to enhance diversity. The proposed SLDCNet employs a customized EfficientNetB3 backbone, fine-tuned with a global average pooling layer, compressed dense layers (512, 256, 128, and 64 neurons), and a 25% dropout mechanism to facilitate domain-specific feature learning. Model performance was evaluated using standard metrics alongside stratified 5-fold cross-validation and benchmarked against established architectures, including ResNet50, MobileNetV2, InceptionV3, and VGG16. Grad-CAM visualizations enhanced model interpretability by highlighting disease-specific regions. The proposed model achieved training, validation, and testing accuracies of 99.98%, 99.40%, and 99.03%, respectively, with 99% precision, recall, and F1-score. This approach reduces economic losses and quality impairment, facilitating timely interventions to sustain Solanaceous vegetable production.