Brazil Nut Fruit Detection Dataset from UAV Imagery

This dataset contains UAV-based RGB imagery of Bertholletia excelsa (Brazil nut) tree canopies collected in a native forest fragment in eastern Acre, Brazil. All images were manually annotated for the “burr” (fruit) class to support object detection and are provided in 640 × 640 pixel JPG format with a YOLO-compatible annotation files (.txt). A DJI Mavic 3 Pro was used to acquire canopy-level imagery of Bertholletia excelsa trees during the dry season, between 08:00 and 11:00, under natural lighting conditions. The dataset includes both resized UAV photographs and additional images obtained through data augmentation. The augmentation procedures applied image rotation, adjustments in sharpness, addition of noise, zoom transformations, and cropping operations. The dataset was originally developed for the study:Pereira de Carvalho, H.; Barros, Q.S.; Ferreira, E.J.L.; Ferreira, L.; Rodrigues, N.M.M.; Silva, L.F. da; Almeida, B.T. de; Gomes Cruz, E.; Pinheiro, R. de M.; Pádua, L. A Dataset for Brazil Nut (Bertholletia Excelsa Bonpl.) Fruit Detection in Native Amazonian Forests Using UAV Imagery. Agronomy 2026, 16, 341, doi:10.3390/agronomy16030341.