Dataset - MusselSeg: Semantic Segmentation for Rocky Intertidal Mussel Habitat

[hakai-it]

MusselSeg: Semantic Segmentation Dataset for Rocky Intertidal Mussel Habitat

https://hakaiinstitute.github.io/hakai-metadata-entry-form/#/en/hakai/DCH8GzsQKSM8xwJ8WdQFIZrtCaq2/-O26fyGiVXoF__M22UrP

Best Practices Checklist

In General

• [x] No previous versions of this metadata record exist (eg for earlier versions of the data, if so update that record rather than creating a new one)

Data Identification

Dataset title:

• [x] No version information in the title
• [x] Frontloaded (with the most important information first)
• [ ] Include the geographical region the data apply to
• [x] Short – aim for 60 characters including spaces
• [x] Does not include acronyms – put these in the keywords
• [x] Does not include the word “dataset”
• [x] Time series datasets should include “time series” at the end of the title

Abstract

• [x] Abbreviations have been expanded upon at first mention
• [x] Abstract describes how, when, what, where, why of data collection and is limited to no more than 500 words

DOI

• [ ] A DOI has been drafted for this record
• [ ] DOI has been updated via the form after review and changes to record
• [ ] DOI has been manually edited on datacite fabrica
• [ ] DOI status has been changed from Draft to Findable

Spatial

• [x] Ensure that Depth or Height Positive is correctly selected

Contact

• [x] ROR and ORCID(s) are included and linked properly where applicable
• [x] For datasets where DFO is a partner, ensure 'parent' ROR is added (https://ror.org/02qa1x782). DFO 'child' organizations (i.e. CHS) and their ROR are optional.
• [ ] Include Hakai Institute as Publisher and include data@hakai.org as email
• [ ] Make sure email address is provided if the role is 'Metadata Custodian' or 'Point of Contact'
• [x] Add contact affiliation where known including ROR
• [x] If resource is (partially) generated by Hakai researchers, include 'Tula Foundation' (with associated ROR) with 'Funder' role. Be sure to uncheck 'include in citation' for Tula Foundation.

Resources

• [ ] Resource links go to specific dataset download (not generic platform like waterproperties.ca)
• [ ] Readme, changelog, data dictionary, protocols included in data-package (for tabular text based data)
• [ ] An archive folder, or other means, for older data versions is included in the data package if the version is not 1.0
• [ ] Links work
• [ ] All files in the data package can be opened and are not corrupt
• [ ] No executable files in the data package. Files should be open formats and standards (.csv, .txt for example)

[JessyBarrette]

@willhakai Thanks for submitting a metadata record for the dataset:

MusselSeg: Semantic Segmentation Dataset for Rocky Intertidal Mussel Habitat

Couple thoughts:

• [x] Would you mind adding your email address to your contact since your the metadata custodian
• [x] Can we add Hakai Institute as data owner
• [x] Can we add Tula as funder
• [x] I would perhaps add a rough region to the title like: MusselSeg: Semantic Segmentation Dataset for North East Pacific Rocky Intertidal Mussel Habitat
• [ ] Should we generate a DOI for the dataset?
• [x] Is it expected the that huggingface dataset isn't accessible yet?

[timvdstap]

I think @tayden is going to add you to the Hakai HuggingFace account Jessy :) - also, I think that #105 is this records' duplicate?

[tayden]

The dataset is located here: https://huggingface.co/datasets/HakaiInstitute/mussel-seg-1024-1024 It's currently private, so you'll have to make an account and I can then add you to see it. It'll be made public once it gets the rubber stamp of approval.

[willhakai]

@tayden is creating a DOI on HuggingFace

Other changes made, thanks!

[tayden]

@tayden is creating a DOI on HuggingFace

Other changes made, thanks!

Yes I will do this. I can't do it until the dataset is public, so I will then. The HF DOIs will track the dataset versions automatically, so it'll be easier to do there rather than with an external service

[fostermh]

hmmm for the DOI are we talking about one that points to the data resource or the metadata record. The normal way we do this is to generate a DOI that points to the metadata record so that we can change the location of data storage if needed. There is a handy button in the form for just this. Or does HF require a doi that points to their site? if so we can do both.

[tayden]

HF has one that points to the dataset, on HF. The main advantage to the HF one is that you can track a specific dataset revision.

That being said, there's no reason I couldn't reference a DOI that links to the metadata record instead, but we'd lose the automatic revision tracking.

[fostermh]

ok, so generating both in this case would make sense then. one on HF that points to their page, another on the form that points to our page, and we indicate that the HF one is identical to the hakai one, which we can do by adding the HF doi under 'related works'.

I realize that from a technical perspective non of this is needed and we could just use the HF doi. The issue being addressed here is one of appropriately representing ownership and credit. Wherever possible we want to link external data/metadata back to hakai records so that we can have an accurate list of hakai data holdings.

Anyway, the short version is Taylor you should carry on generating a doi on HF and link it to HF then add it to the metadata record under related works.

Thank you.

[tayden]

Thanks @fostermh!

[tayden]

• Does not include the word “dataset”

I've updated this and removed "dataset"

[JessyBarrette]

Ok I cleaned up the contacts for hakai/tula to not have duplicated names.

We can certainly add a DOI to this specific repo, you can then still generate a DOI on hugginface and reference your version specific DOI for any publications. We can ourself tracks related DOIs external to the organization.