### The dataset includes:

- Training data: 1000 images, segmentation ground truths, and the corresponding run-length encoding.

- Testing data: 300 images.