A Novel Deep Learning Framework for Cactus Species Classification

The identification of cactus species depended on manual assessment conducted by professionals. The manual evaluation is laborious and inconsistent because of the nuanced differences in morphological traits. This study presents a novel framework for an automatic classification system of Astrophytum asterias var. nudum utilizing VGG19 and ResNet152V2. We gathered cactus data from social media and a cactus farm in Thailand. The data labels received expert approval. The optimal model for identifying the normal and special classes was VGG19, which achieved a superior F1-score of 0.867. Additionally, ResNet152V2 achieved an F1-score of 0.78 in recognizing the various special classes.