uploading datasets
silburt opened this issue · 5 comments
I think that the train/dev/test datasets belong on zenodo (and provide a shareable link in the DeepMoon readme and paper), but I think a final CNN-predicted crater distribution (for train/dev/test) belongs in this repo (in csv format), in the same folder as our LROC and Head datasets. What do you think @cczhu?
Ultimately up to you, but I don't see a good reason to do so, unless you're uploading it so people can compare its accuracy against the Head and LROC sets. Given that we have a of-order 0.1 degree long/lat offset and a couple of percent offset in crater diameters, it can't be used as an actual crater catalogue (nor was that ever our intention, as Kristen mentioned), so we should definitely flag it or highlight that in some way in case people accidentally use it as an input catalogue.
I think it would be nice to put it somewhere for people to look at/play around with, along with a disclaimer that it's not intended to be used as a ground truth catalogue. Maybe I'll put it in the zenodo DOI then. Speaking of that, is there anything else you want in the zenodo link? Right now I have the image/crater catalogs, the best model, and now maybe the final crater distribution. What about the Global DEM? Right now in the README.md is says "XXXXXX".
I do think the Global DEM is a good idea (because a .png version needs to be converted from .shp off the USGS site). I also think it's a good idea to include a Readme.txt that describes the origin of each file.
@cczhu I've uploaded the test/train/dev_craters.hdf5 so far to zenodo, but the images are painfully slow on my end (I'm using the sshfs route). Any chance you have an efficient option on your end to upload them? Otherwise I suppose I'll try and download them first to my local machine and then upload them to zenodo...