An Ensemble-based Deep Learning Model for Multi-class Flower Recognition

Many species around the globe heavily depend on flowers, and even the use of flowers for medical purposes is enormous, which has been in practice since ancient times. In modern botany, agronomy, and species research, the identification and classification of flowers mainly depend on human intervention, which requires a lot of human resources, and the recognition accuracy rate is low. On the other hand, due to the vast number of flowers with different types, it is quite a challenge and increases the complexity of the accurate detection and classification process. Moreover, the traditional CNN model and other pre-trained models like- VGG16, VGG19, Xception, and MobileNet-V2 failed to ensure high validation accuracy. In this research, we have designed an ensemble-based deep learning model consisting of two pre-trained models DenseNet-201 and EfficientNet-B7, which we renamed “FlowerConvNet”. This paper considers classifying five categories of flowers consisting of 4242 images, collected from the Kaggle flower recognition dataset. In this research study, the proposed deep learning model FlowerConvNet using transfer learning can obtain a competitive result with 95% validation accuracy.