Classification of mixed color rose types using convolutional neural network

Abstract

Roses are one type of thorny flower that has a beautiful smell alongside ornamental attributes. Roses have more than 150 species, many of which can be differentiated by different colors and sizes. Ordinary people with limited botanical knowledge usually can't classify flowers just by visually observing them. Additionally, even if they could, it would be difficult to classify and mark all the flower species due to the sheer volume of the amount in a garden. In this particular work, we have designed a system that can classify the type of rose automatically through the usage of machine learning. Using such an autonomous tool, rose species can be identified without any manual intervention, which can be useful in many applications. For sample representation, we have classified six different rose species such as pink-red rose, rainbow rose, Neil Diamond, Graham Thomas, Abracadabra ROSE, and Double Delight Rose. This is done through the application of a deep learning framework with a convolutional neural network model. Experimental results show that the model can achieve 85% validation accuracy in the task of classifying six rose species, with satisfactory performance in classifying each individual species. In the next phase of the experiment, we performed an identical analysis on the grayscale version of the dataset to showcase the significance of the color feature in identifying rose species.