CLASSIFICAÇÃO DE ESPÉCIES DE ABELHAS NATIVAS SEM FERRÃO UTILIZANDO REDES NEURAIS CONVOLUCIONAIS

Abstract

Bees are essential for pollination in agriculture and, consequently, for food production. However, manual identification of these species can be inaccurate and cause stress to the insects, making it essential to develop automated methods for this task. This study aims to use Convolutional Neural Networks (CNNs) to recognize and classify bee species quickly and accurately, eliminating the need for manual identification. To this end, a balanced database was built on the Roboflow platform, divided into five classes, each representing a bee species. The CNN model was trained using this dataset, allowing the application to identify the species from images submitted by users. After training, the CNN presented an accuracy of 79\% with a relatively small dataset, demonstrating promising performance and outperforming conventional methods that are inaccurate and time-consuming. The results obtained indicate that the CNN can replace manual methods of bee identification and classification, offering a more efficient and reliable process. This advance can contribute to the preservation of pollinators and sustainability in agricultural production.