Trying to train a semantic segmentation network to segment clouds from the satellite images, but when I open the image by the rasterio, I find out many images are dark, the data owner told me those data are preprocessed by QGIS, they change the contrast manually and draw the mask. Problem is, I don’t know what kind of preprocessing the owners did when they creating the masks.
Is there a recommended solution to preprocess these kind of images? I simply divide the rgb bands by the maximum value
raw_rgb / np.iinfo(raw_rgb.dtype).max
Any suggestions, thanks.