Year: 2025
Type: dissertation
Abstract: This work aimed to develop a dataset and a classification model for identifying Ipê trees using supervised machine learning techniques. The methodology involved automated image collection through web scraping, dataset curation and organization, and feature extraction using the Res- Net50 network. Since the problem was formulated as a binary classification task-distinguishing between images of "ipe" and "non-ipê"-different classifiers were tested, including Random Fo- rest, Decision Tree, Logistic Regression, and K-Nearest Neighbors. However, the Support Vector Machines (SVM) classifier achieved the best overall performance. The results demonstrate the potential of this approach to support environmental mapping studies and the automatic identifi- cation of species.
Author: Melo, Guilherme Augusto Silva de
Language: Portuguese
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7155455994
