Hybrid CNN-Ensemble Framework for Robust Floral Species Recognition

This study introduces a sophisticated floral identification system based on Deep Learning and Machine Learning to improve species classification accuracy. The system combines VGG16 CNN for feature extraction and fine-tuning classification with a weighted voting ensemble of Random Forest and SVM. Data augmentation techniques such as rotation, brightness, and random shifts were employed to improve model flexibility and generalization. These augmentations significantly enhance the precision of classification, giving robust and reliable identification of species. The model has been developed by TensorFlow, Python, and Scikit-learn, which enables smooth training, tuning, and test operation. Experimental evaluation shows marked improvement in precision, recall, and F1-score, asserting the efficacy of the proposed approach. The use of the weighted ensemble method provides further improvement of classification consistency with balanced predictions made by various classifiers. The findings establish the system as a viable mechanism for automated floristic species classification, which becomes a useful aid in botanical exploration, conservation biology, and in agriculture. Through this study, the benefits of integrating deep learning with conventional classifiers to develop a highly scalable yet accurate floral system are established, achieving an overall accuracy of 90.3%.