FUSU

FUSU dataset covers 5 whole urban areas, 847 km^2 located in the north and south of China, with 17 land use and land cover (LULC) classes and over 170K images and 30 billion pixels of annotations, supporting segmentation, change detection and domain adaptation tasks. This data comprises 2 parts:

Bi-temporal high-resolution satellite RGB images with fine-grained annotations.

Monthly revisited Sentinel-2 and Sentinel-1 images.

Example: T1: im1/6_255.png, im1_label/6_255.png T2: im2/6_255.png, im2_label/6_255.png Sentinel: A/{8-12}/6_255_A_{8-12}.png, B/{1-12}/6_255_B_{1-12}.png, C/{1-12}/6_255_C_{1-12}.png

{A,B,C} represents the year, and {1-12} represents the month.

Shape: T1 image (512 times 512 times 3), T2 image (512 times 512 times 3), T1 label (512 times 512 times 1), T2 label (512 times 512 times 1), Sentinel image (128 times 128 times 14).

14 bands of Sentinel images include 12 bands for Sentinel-2 and 2 bands (VV/VH) for Sentinel-1. Now they are concatenated, we will separate them into single .npy file in the future for easier usage.