Closed amardeepjaiman closed 5 years ago
Thank you for pointing this issue! You're totally right, that is a urgent point to address.
We should modify the way aerial
datasets are preprocessed, so as to make them similar to tanzania
dataset. That's looks related to #91. I add this to my todo list.
Hi,
In extract_coordinates_from_filenames(filenames): function, coordinates cant be decoded from tile_imagename using below code. basenames = [os.path.splitext(os.path.basename(f))[0] for f in filenames] coordinates = [b.split("")[-2:] for b in basenames] return [[int(x), int(y)] for x, y in coordinates]
For aerial dataset, the tile names are "austin14_10.png". the above code can not work with this naming convention. I think this code was written for Tanzania dataset. Can someone guide how to get x,y value from this naming convention ?
Thanks, Amardeep