Replication Data for: Recognition of urbanized areas in UAV-derived very-high-resolution visible-light imagery

The attached images contain fragments of orthomosaics and reference classification into urbanized and non-urbanized areas. The pixels classified as urbanized areas contain objects that do not change their appearance/texture over a long period and are fixed to the ground. Pixels classified as non-urbanized areas contain all other objects—mostly bare soil and vegetation. Datasets from two study areas and several survey series are included. Data from Jerzmanowice, Poland (JERZ) were collected in 11 survey series and data from Wieliczka, Poland (WIEL) were collected in four survey series. For these areas and each survey series, the UAV-derived orthomosaics were created with a spatial resolution of 10 mm/px for JERZ and 15 mm/px for WIEL. Then, for each orthomosaic 54 square sample areas with an area of 4 m2 (201 px x 201 px for JERZ and 134 px x 134 px for WIEL) were systematically selected in a uniform grid (random sample). For each pixel in these samples and each analyzed survey series, a reference classification was performed manually. For each dataset (JERZ, WIEL), sample areas from one series were combined in one row with a height of 201 px or 134 px, respectively. Then, data from subsequent series were combined vertically. The order of sample areas in all series is the same. There are therefore two images available for each dataset, i.e. combined fragments of orthomosaics (sample areas) and the corresponding reference classification. Data from orthomosaics are presented in RGB space. In images with a reference classification, 0 corresponds to a non-urbanized area, and 1 corresponds to an urbanized area. For the JERZ dataset, the files contain data for 54 sample areas observed in 11 series in the period from 2021.09 to 2022.08. The order of the series in the composite image is as follows (from top to bottom direction): 2021.09, 2021.10, 2021.11, 2021.12, 2022.02, 2022.03, 2022.04, 2022.05, 2022.06, 2022.07, 2022.08. The file does not include the data from 2022.01 series as it was removed due to snow cover making it impossible to perform correct classification based on RGB images. For the WIEL dataset, the files contain data for 54 sample areas observed in four series in the period from 2021.11 to 2022.08. The order of the series in the composite image is as follows (from top to bottom direction): 2021.11, 2022.02, 2022.05, 2022.08.