Assessing the Taxonomic Accuracy of Mobile Plant Identification Applications among Tropical Plant Species

Abstract

Plant identification through mobile apps has gradually grown in popularity over time. With the emergence of a variety of plant identification applications, problems arise in terms of misidentification of plant species through these applications. Hence, this research aims to assess the accuracy of mobile applications designed for identifying tropical plant species, which are essential for biodiversity research and conservation in tropical regions. The study tested five selected tropical plant identification applications, in particular PlantNet, PlantSnap, iNaturalist, PictureThis, and LeafSnap, against a reliable botanical database at the Penang Botanical Garden, using a varied sample of 100 plant taxa, including trees, shrubs, and herbaceous plants. The evaluation of the applications' accuracy was conducted at three taxonomic levels: species, genus, and family. The results indicated that PlantNet achieved the highest accuracy at the species level, while iNaturalist performed best at the genus level. The proposed procedure emphasizes key factors including lighting conditions, image focus, viewing angles, plant morphological characteristics, and the inclusion of relevant metadata. The findings highlight the potential of mobile plant identification applications as valuable tools for biodiversity documentation and conservation, while also emphasizing the importance of proper image acquisition techniques to enhance identification reliability.