Efficient Indian Flowers Recognition Model Using Deep Convolutional Neural Networks

The objective of this study is to create an effective recognition model utilizing deep Convolutional Neural Networks (CNN) for identifying 24 distinct varieties of Indian flowers. The dataset employed in this investigation comprises 8066 raw images representing the 24 species of Indian flowers, with 350 images available for each flower type. A sequential CNN model is employed to assess the CNN's performance on images of the 24 flower types. The findings reveal that Convolutional Neural Networks exhibit favorable outcomes for the limited task of flower recognition. However, the results are susceptible to factors such as image type, image quality, the number of epochs, and the architectural depth of the CNN. The models exhibit satisfactory performance metrics, achieving the highest training accuracy of 85.60% and validation accuracy of 61.99%, respectively. Moreover, this developed model holds potential for recognizing flower species beyond the Indian context, making it applicable in other regions globally. Additionally, the research produces a pre-trained model that can be effectively applied to identify various types of flowers beyond those included in the initial training dataset.