FAIR1M Dataset labels_root name

[SpontaneousDuck]

Description

I just downloaded the FAIR1M dataset and found that the labelXmls.zip files extract to a folder named labelXml instead of the folder expected by torchgeo.datasets.FAIR1M shown below.

torchgeo/torchgeo/datasets/fair1m.py

Line 160 in d897e33

labels_root: str = "labelXmls"

I can get around this by extracting to the expected folder myself but this is not how the zip extracts by default. Did this maybe change recently in the dataset?

Steps to reproduce

1. Download FAIR1M dataset from the Google Drive on their website (https://www.gaofen-challenge.com/benchmark)
2. Create dataset object

from torchgeo.datasets import FAIR1M
ds = FAIR1M(root='/data/FAIR1M/train/part2')

3. Check data root and see the folder name mismatch. This occurs in all the data folders.

$ ls /data/FAIR1M/train/part2
images  images.zip  labelXml  labelXmls.zip
$ ls /data/FAIR1M/train/part1
images  images.zip  labelXml  labelXmls.zip
$ ls /data/FAIR1M/validation
images  images.zip  labelXml  labelXmls.zip

Version

0.4.0