Flower Phenotype Recognition and Analysis using YoloV5 Models

Abstract

Real time detection of flowers based on their growth cycle is called as phenotype. It is one of the most important methods for judging the maturity of flowers and to estimate their yield. Traditional method involves flower detection and classification of varied species. In this paper, we introduce a new dataset based on flower phenotype. The dataset has images of flowers, classified into bud, fresh and stale. The work will help for identification and localization of classes based on flower phenotype. The detection of flowers at various stages of their life will be more important to harvesting in floriculture field. We propose a state of art deep learning-based approach using YOLOV5 model is used for identifying flowers based on flower phenotype. The images are subsequently augmented using rotation transformation, color balance transformation, brightness transformation and blur processing. The augmented images are used for preparing training sets. Using different versions of YoloV5 the flower image dataset is trained and tested. Primarily, flower phenotype considered has stages like bud formation, blossoming and stale flower. © Grenze Scientific Society, 2022.