I am working with rle-encoded masks. There is just -1
if there is no mask and a string containing an rle encoded mask if there are annotations.
But I do not have any labels to the masks. If I try to create a databunch from the data I get the error:
Exception: Your data isn't labeled, can't turn it into a
DataBunchyet!
Why do I need labels?
This is how I try to create my databunch:
src = (SegmentationItemList
.from_csv(path_jpgs, 'train-rle-sample.csv', header=None, convert_mode='grey')
.split_by_rand_pct(0.3))
data = (src
.databunch(bs=bs)
.normalize(imagenet_stats))