Deep Learning for Plant Species Classification

Abstract

Neither the ecology of the world nor the existence of humans can exist without plants. It is necessary to have an automated technique that makes use of deep learning in order to safeguard endangered species. In order to pre-process leaf images and extract important features, a new CNN-based technique known as D-Leaf was introduced. Specifically, this approach takes use of three distinct CNN models: D-Leaf, pre-trained AlexNet, and fine-tuned AlexNet. The support vector machine (SVM), artificial neural network (ANN), k-nearest neighbour (k-NN), naive bayes, and convolutional neural network (CNN) were the five various machine learning approaches that were used in order to establish the classification of these qualities. The D-Leaf model surpassed both the raw AlexNet model (93.26% accuracy) and the fine-tuned AlexNet model (95.54% accuracy) in the testing process. Additionally, the ANN classifier was an excellent choice for the CNN that was emphasised. According to the findings of the empirical research, D-Leaf has the potential to serve as an effective automated method for the classification of plant species.