estherrolf / implicit-posterior Goto Github PK

Hi there,
Thank you so much for releasing the code.
I'm wondering how you dealt with the date inconsistency between OSM data (for roads and water), NAIP imagery, and high-res labels. For example, Phoenix high-res labels (from EPA's MULC EnviroAtlas) are obtained according to NAIP imagery from 2010. On the other hand, OSM data from Osmnx is up-to-date. The same problem exists for Microsoft Building Footprints (the earliest available data is from 2017).
I'm curious what you think about this type of discrepancy and how you handled it when preparing the EnviroAtlas dataset. Do you have any suggestions to deal with this problem?

Thanks,
Sorour

Hello, your article is very helpful. I have some problems in the reproduction process, I hope you could help me. I want to get the prior of the EnviroAtlas dataset, through make_priors_envirotlas.py. But I don't know how to get the csv file in ea_data_splits_dir = '/home/esther/qr_for_landcover/enviroatlas_data_splits'. Looking forward to your reply, thank you!

Hello,
I greatly enjoyed your paper, an am interested in understanding the approach of ranked based learning described in section 4.2. I see code here for the reconstruction of the negative labels with MNIST but I do not see anything for rank based learning on MNIST. Could you direct me to the relevant script?