Comparing Convolution Neural Network models for leaf recognition

Abstract

This research compares the recognition performance between pretrained models, GoogLeNet and AlexNet, with basic Convolution Neural Network (CNN) for leaf recognition. Lately, CNN has gained a lot of interest in image processing applications. Numerous pretrained models have been introduced and the most popular pretrained models are GoogLeNet and AlexNet. Each model has its own layers of convolution and computational complexity. A great success has been achieved using these classification models in computer vision and this research investigates their performances for leaf recognition using MalayaKew (MK), an open access leaf dataset. GoogLeNet achieves a perfect 100% accuracy, outperforms both AlexNet and basic CNN. On the other hand, the processing time for GoogLeNet is longer compared to the other models due to the high number of layers in its architecture.