This dataset contains over one thousand high-resolution images of aerial photographs taken in southern California. The photographs were taken from a plane and are meant as a reasonable facsimile for satellite images.
Images are grouped into sets of five, each of which have the same setId. Each image in a set was taken on a different day (but not necessarily at the same time each day). The images for each set cover approximately the same area but are not exactly aligned. Images are named according to the convention setId_day. In the training set, the images have the correct day ordering. In the test set, the images are scrambled.
This competition asks you order the scrambled images in the test set. This is anticipated to be a challenging task. Some locations will provide very little evidence of changes from day to day. Other locations will have subtle changes that hint at an ordering, such as parked or moving vehicles, differences in bodies of water, or changes in shadows. On account of the difficulty, you may use both manual and computer-based analysis in this competition.
For convenience, we have provided both lossless, high-quality images (train/test) as well as the same images in a compressed jpeg format (train_sm/test_sm).