Reconhecimento de espécies florestais pela folha, utilizando redes neurais convolucionais

Abstract

Seeking the control and preservation of plant species, the identification stage becomes a primordial activity. The identification process is carried out mainly by botanists, where the identification occurs with the comparison of already known specimens or with the aid of books, manuals or identification keys. Artificial neural networks have been shown to be a good accuracy for classification problems, so the project used convolutional neural networks to classify species by leaf image. In the project, 29 species were collected at the Universidade Tecnol´ogica Federal do Paran´a at the Medianeira campus. Two network models, YOLO and Googlenet, were used due to the difference in image manipulation and a comparison between them. The YOLO and Googlenet models achieved recognition of 86.2 % and 90.3 %, respectively.