A dataset of high-resolution UAV images of floods and waterlogging in Jiujiang, Jiangxi Province

The dataset consists of two data folders, "origin" and "clip", and three text files: "train.txt", "val.txt", and "test.txt". The "origin" folder contains two subfolders, "images" and "targets", which store the original UAV high-resolution images and their corresponding annotation images in JPG and PNG formats, respectively, with a total of 412 samples. The "clip" folder contains non-overlapping cropped and selected 512x512 pixel images and their labels, stored in the "images_cropped" and "targets_cropped" directories, with a total of 12,080 samples. The three text files contain the names of the training, validation, and test samples, randomly shuffled in an 8:1:1 ratio.