-
Notifications
You must be signed in to change notification settings - Fork 8
Add WorldCover segmentation task #274
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
base: master
Are you sure you want to change the base?
Conversation
Farbum
left a comment
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.
LGTM.
Does it make sense to add a readme.md in the worldcover folder to refer to the tolbi readme.md for data spec and collection information?
| average: "micro" | ||
| other_metrics: | ||
| cacao_precision: | ||
| class_path: rslearn.train.tasks.segmentation.SegmentationMetric |
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.
Would it make sense to set report_metric_per_class to True in the SegmentationTask init parameters to have metrics reported per class?
Never mind, we just talked. We'll need to add a naming dictionary to the logic.
| @@ -0,0 +1,214 @@ | |||
| """Create windows for the Tolbi project.""" | |||
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.
Fix docstring (Tolbi -> WorldCover) and add some information either in docstring or README about where the WorldCover data is, how much data there is, etc.
| COPY requirements-extra.txt /opt/rslearn_projects/requirements-extra.txt | ||
| # Using cache mount here avoids needing to re-download dependencies for later builds if the version didn't change. | ||
| RUN --mount=type=cache,target=/root/.cache/uv uv pip install --system /opt/rslearn[extra] /opt/olmoearth_pretrain -r /opt/rslearn_projects/requirements.txt -r /opt/rslearn_projects/requirements-extra.txt | ||
| RUN uv pip install --system /opt/rslearn[extra] /opt/olmoearth_pretrain -r /opt/rslearn_projects/requirements.txt -r /opt/rslearn_projects/requirements-extra.txt |
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.
What is the reason to drop this? I thought it makes the build faster.
This PR include two tasks: (1) Tolbi cash crop classification - though given the lack of negative samples, the inference result is not good, (2) WorldCover land cover and land use classification - this is using the labels here: https://doi.org/10.5281/zenodo.14871659, and tried different modality combination (i haven't ran inference yet, will do that later).