Training data semantic segmentation historical Images

This dataset contains historical images and their labelled counterparts used to train a network for semantic segmentation of historical images. The images are historical aerial imagery from Antarctica, made between 1940 and 1990.The images are black and white and can be nadir (V) or oblique (L or R).The labelled images are labelled manually and are the segmented equivalents of the historical images. They contain 6 classes:1: ice, 2: snow, 3: rocks, 4: water, 5: clouds, 6:sky, 7: unknown.A pixel value of 0 means unlabelled.<br>