Customized EfficientNet with Squeeze and Excitation model for classifying medical blossoms

This research introduces a customised architecture of EfficientNetB10, integrated with Squeeze-and-Excitation (SE) Blocks and Swish activation for enhanced feature extraction, sharpening an advanced deep learning model for flower classification. Syringa, Bombax Malabarica, and twelve other species formed the dataset used for training the model. The dataset underwent robust preprocessing steps including normalization, resizing, and augmentation. The architecture employed for feature extraction was based on EfficientNetB10, adjusting the last twenty layers' unfrozen parameters to ensure proper flower image adaptation.