-
Notifications
You must be signed in to change notification settings - Fork 378
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add South Africa Crop Type Competition Dataset #1840
Add South Africa Crop Type Competition Dataset #1840
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
* Documentation, satellite-specific transform and weights for additional Satlas single-image rgb&multispectral Swin-v2 models. Tests pass. * Address 3 of comments * Address comments, fix readmydocs and isort, mypy still unhappy * update * Add bands to meta dicts * Add comment about Satlas S2 RGB using TCI product * linting --------- Co-authored-by: Piper Wolters <piperw@allenai.org> Co-authored-by: Piper Wolters <piperw@prior-cirrascale-79.reviz.ai2.in>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
…eHuber/torchgeo into add-south-africa-crop-type
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
…eHuber/torchgeo into add-south-africa-crop-type
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don't see any remaining major issues. Just wondering if __getitem__
can be further simplified. Also need comments in __getitem__
to explain all the non-standard stuff.
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
I added some comments to explain the extra stuff, let me know if you think it needs more detail. The multiple images to one label thing is really where most of the complication comes from along with the need to make sure the labels and fields are in the same order since they won't be sorted alphabetically. I thought this would be the most efficient solution as we loop over the hit filepaths once just as we would standardly but I'm open to other ideas. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Probably worth iterating over the real data just to make sure there are no bugs once you move away from the synthetic data. Just as a quick personal test, doesn't need to be in CI.
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
Returns: | ||
data and labels at that index | ||
""" | ||
assert isinstance(self.paths, str) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It's probably possible to relax this constraint by using filepaths
to compute the correct directory for the mask. But let's do that another day, I want to get this merged.
This PR adds the South Africa Crop Type Competition Dataset to TorchGeo.
Dataset
Similar to #1459, this dataset does not implement automatic download functions due to issues migrating data from radiantmlhub source cooperative until issue #1830 is resolved.
Example field and label in early July: