Failure of matching datasets of WMH

[lorenzoyuan]

Hello, I've downloaded the datasets of WMH proposed in the website https://doi.org/10.34894/AECRSD, but I failed to match it with the code of boundary-loss in my server. It was written that "md5sum: WARNING: 3 computed checksums did NOT match", and there wasn't a folder named "Amsterdam_GE3T". What should I do ?

[HKervadec]

(Sorry for the delayed reply. Somehow the Github notification for this repo are still broken on my side.)

You seem to have linked to the "latest" release of WMH, the 1.0 (it was made public recently, and included the originally private test set). Though now I notice that http://wmh.isi.uu.nl/data/ went 404.

I had that chat with Hugo sometime ago, the Dataverse website seems to generate .zip on the fly, so that checking their integrity is not feasible. So I opted for another solution on a different code-base.

For now, you could disable the data check, and would need to modify the slice_wmh.py script for the new release.

I have a version running for WMH1.0 running on a different code-base, I might be able to back-port it to this repo in the coming weeks.