Eficácia de diferentes aplicativos para smartphones na identificação de espécies de árvores nativas usadas na restauração da Mata Atlântica

Abstract

Botanical identification is extremely important in various stages of ecological restoration, whether in complying with legal requirements, ecological monitoring, or ensuring the success of restoration actions. Many methods are available for this purpose, and more recently, the use of artificial intelligence (AI) in species recognition has become prominent, such as in smartphone applications. This study evaluated the effectiveness of the PlantNet, iNaturalist, and Google Lens applications in identifying native tree species used in Atlantic Forest restoration. Leaves, flowers, and fruits were photographed and submitted to the applications under standardized conditions. The results indicated better performance by iNaturalist, followed by PlantNet, both with high accuracy rates on the first suggestion and low error rates, while Google Lens showed lower reliability, with a high frequency of incorrect suggestions. Greater accuracy was observed when using images of flowers, although this condition is not always available. The use of artificial backgrounds in the photographs decreased accuracy. In conclusion, although specialized applications represent promising tools to support identification activities in ecological restoration, their isolated use does not replace botanical knowledge, and validation of the suggestions is necessary to avoid errors that could compromise properly identification of restoration stages.