2 Datasets of forests for Computer Vision and Deep Learning techniques

Datasets:

1) Coastal forest in Shonai area; contains an orthomosaic (processed by Metashape) of the coastal forest and annotated layers (processed in Gimp)

Annotated layers: 0 = annotations of black locust; 1 = annotations of soil; 2 = annotations of man-made; 3 = annotations of other trees and the orthomosaic as JPEG file

2) Mixed forest images of the Yamagata University Research Forest taken in the winter season; contains 7 orthomosaics (TIFF file), annotated layers (named wM1 to wM7; processed in Gimp)

Orthomosaics: 7 orthomosaics of 5 different sites; for one of the sites we provide 3 orthomosaics of different days and illumination conditions

Annotated layers: 0 = annotations of river class; 1 = annotations of deciduous class; 2 = annotations of uncovered class; 3 = annotations of evergreen class; 4 = annotations of man-made class and the orthomosaic as JPEG file

The dataset was used for our transfer learning study in the field of forest applications. We used in one experiment the winter images to run deep learning algorithms. In a second experiment we used the coastal forest data for a similar approach.