search

sfg-taxonpages / orthoptera

0 stars 0 forks source link

Metadata of pictures, particularly after download #14

Open klausriede opened 10 months ago

klausriede commented 10 months ago

pictures canbe downloaded, this is good! However within the downloaded picture I don´t find any info in the EXIF about source, copyright etc grafik

Eventually it might make sense to embed this info (copyright, metadata) inside the picture at the lower edge. A crypto stenocode might also be a good idea... here you find a screenshot of the downloaded picture . No author info, no source...This is not FAIR ... grafik

klausriede commented 10 months ago

sorry here comes the starting point Phyllomimus klapperichi grafik

mjy commented 10 months ago

@klausriede thanks for the insights. From my knowledge there is no one way to be FAIR, and indeed while the principals are commonly espoused the practice is much more of an art than a science, at least that's my experience. For example, do major providers like iNat or WIkiCommons synchronize the (sometimes vast) metadata on the image with the EXIF (note that not all image times even support this format)? What happens when metadata change? What should we do with bad EXIF data "out there" and no longer in control of our curatorial process? We can raise many questions as to whether its reasonable to do what you propose, or even useful in the long run.

There is a perhaps more robust way to think about this. We track the SHA of all these images, the unique fingerprint that only this binary blob can possess. We offer many FAIR endpoints on TP for each and every page (see SiteMap, e.g. https://sandcastle.taxonworks.org/api/v1/otus/44862/inventory/images.json?project_token=NOw8ZKe7O0axmMkNPZn35Q&extend[]=depictions&extend[]=attribution&extend[]=source&extend[]=citations&otu_scope[]=all) that provides you all the basic metadata in machine readable form. We offer many many additional discovery points in the TaxonWorks API endpoints. Using the fingerprint attribute we can then provide data that reference this "Thing" in a very robust way, moving forward and taking this data into other forms.

So, in general, we're still evolving, but relative to our past efforts I think we might argue we've made some nice progress?