This dataset is composed of Sentinel-1 and Sentinel-2 data of several areas of the Brazilian Cerrado for the development of supervised deep learning algorithms on SAR-optical data fusion. The reference data created was built based on the TerraClass Cerrado 2018 map of the same area and the vegetation map produced by Bendini, Hugo N; Fonseca, Leila Maria Garcia; Schwieder, Marcel; Rufin, Philippe; Körting, Thales Sehn; Koumrouyan, Adriana; Hostert, Patrick (2020): Combining environmental and landsat analysis ready data for vegetation mapping: a case study in the Brazilian Savanna biome. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLIII-B3-2020, 953-960, https://doi.org/10.5194/isprs-archives-XLIII-B3-2020-953-2020.It is composed of 15002 training patches, 4000 validation patches, both of size 384x384, and 10 testing images of aprox. 100x100 square km. The Sentinel-2 data is identified by the S2 prefix, and is composed of the 10m and 20m bands of the original Sentinel-2 image. The Sentinel-1 data is identified as S1 and it has 7 bands in a 10m resolution-cell, consisting of the intensities in both polarizations (VV and VH), the real and imaginary part of the off-diagonal of the coherence matrix, and the 3 elements of the H-alpha dual-pol decomposition. The reference data is presented in two ways: one at 10m, which is identified by the GT prefix, and another at 30m, which is identified by the prefix MN. The 30m reference comes from the original reference maps the dataset was based on, since they were produced by Landsat images.Due to the size of the dataset, the associated TIFF files are stored separately on secure storage at DANS. Please contact DANS via info@dans.knaw.nl in order to obtain the full dataset.

Date Submitted: 2023-09-15