Evaluation of faster R-CNN for flower phenotyping analysis

The study of Flower phenotypes enables Flower observation for improved agricultural management techniques. The problems of variable lighting, occlusions, and morphological differences among flowers hinder the scalability and accuracy of the traditional phenotyping methods in observing flower development over time.The research measures how well Faster R-CNN addresses these detection difficulties to enhance flower tracking capabilities in multi-view image sequences. The proposed solution involves a multi-view analysis framework to capture phenotypic traits from different perspectives, a detection model that can work well in complex scenes with different lighting conditions, and a temporal analysis of the flower growth at different stages. The model attains a mean IoU of 0.84 and a mean confidence score of 0.8304 showing good localization and classification performance of the algorithm. The research develops both Flower detection and localization technologies which helps to establish potential for real-time application multispectral evaluation systems for better plant health assessment along with agricultural practices.