Performance evaluation of the multiclass classification of flowers on diverse datasets

The accurate classification of flowers plays an important role in our life and researches. Due to the high variation of types and species of flowers, the classification is a challenging task. So far, several classification methods have been proposed. However, the accuracy of the flower classification still needs to be investigated and improved. The paper presents the performance evaluation of flower classification. Both handcrafted feature extraction, traditional machine learning classifiers (e.g., k Nearest Neighbors, Support Vector Machine and Random Forest) and deep neural networks (e.g., Alexnet, Resnet, Inception and Densenet) have been employed to improve the accuracy of the classification. The strategies of using the data augmentation are applied to improve the performance of the classification. The classification methods have been evaluated on four public datasets (Oxford-17, Oxford-102, Kaggle-5 and Iris flower datasets) to compare and analyse the strengths and weaknesses.