Improvements of Flower Classification Using Deep Neural Networks with Various Solver Optimization Algorithms

Flowers hold significant importance across various fields due to their multifaceted roles and impact, such as agriculture manufacturing and pharmacological research. Therefore, the precise classification of flowers has extensive implications in real life and has attracted more and more researchers. However, the accuracy of the classifiers still needs to be improved. In this work, we present an enhancement in the pipeline of flower image classification, where some state-of-the-art models are performed and compared. In our proposed process, we fine-tuned various Deep neural network models (DNNs) to classify flower images. To improve the accuracy of the classifier, various optimization algorithms are performed and evaluated. We also implemented some traditional approaches to compare the efficiency of classification methods. These methods are evaluated on four public datasets (Oxford-17, Oxford-102, Kaggle-5, and Iris flower datasets). The experimental results showed that the Densenet-121 network combined with the RMSProp optimizer had a competitive advantage in terms of accuracy compared with traditional approaches.