My model is capturing all kinds of small unpaved roads in rural regions. I'm afraid it will be considered as False positive which will reduce my leaderboard score. Can you confirm that you too are capturing such roads?
Posted by: nikhalster @ May 2, 2018, 11:02 a.m.As was stated in the dataset description, the labels are not perfect due to the cost for annotating segmentation mask, specially in rural regions. In addition, we intentionally didn't annotate small roads within farmlands.
On the other hand, the testing dataset and the training dataset were labeled in a similar manner, so they're most likely consistent.
Posted by: jinghuang @ May 2, 2018, 5:56 p.m.