Open KaiWaldrant opened 1 year ago
The name and summary can be task-specific. See e.g. zebrafish_random
va zebrafish_labs
in label projection.
Organism is generic but currently not required everywhere; however, since it's always going to be available, I'm okay with moving it to the loader and making it mandatory there.
The task-specific datasets have metadata (link) that are not included in the data loaders (link). In v2, we are currently using an additional yaml (link) to add metadata by default. Is it okay if we add the following mandatory fields to the
util.loader
decorator: dataset_name, dataset_organism, dataset_summary?The dataset_id can be derived from the function name, while dataset_description can be derived from the function docstring.
If you agree, I can make a PR for this.