Issues with loaded segmentation data size and labels

[kadanes]

I used voc2012.read_all_data_and_save() to load the segmentation dataset. But after printing the length of voc2012.train_images and voc2012.val_images I get 1464 and 1449. This number seems to be way smaller than the actual data size.

Also, the number of unique labels is 20. Should it not be 21? 1 for background = 255?

np.unique(voc2012.train_labels) gives this:

array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20], dtype=uint8)