Hi, I’m using flukebook for killer whale matching and have a few questions:
if I upload, for example, 30 photos of the same MarkedIndividual.individualID, with a mix of L & R side photos, is there a way to export info on how many of each side of the animal I uploaded? (I assume flukebook automatically assigns a side and tracks this info)
Can I give whales birth and death years, so that Flukebook doesn’t propose whales as matches outside of the years they were alive?
Can I specify to only match against a certain population? E.g. for populations that co-occur in the same Encounter.locationID (e.g. northern resident killer whales and southern resident killer whales in British Columbia), is there a way to only search for matches within one of those two populations, in cases where the population is known?
How does flukebook understand change over time? i.e. does the algorithm understand that nicks can be added but not removed, and that similarity to a photo taken close in time should be weighted more heavily (e.g. for a fin that has changed but many older and a couple recent photos of the individual are in the database, will the software know to still consider a match to be good if it is similar to the couple recent but not the many old photos?)
No, not to that kind of granularity. Right now, it matches against all encounters of the same species within your selected geographical area. This is another good candidate for a feature request.
I’m going to defer to @jason for this one, but broadly, our documentation covers how our image analysis tools work.
to add further detail on my first question, the thing that would be most helpful to me is having a way to check whether I have both a L and R photo for all animals I’ve uploaded, so I can fill in any gaps of animals missing one side.
Yes, in the Search menu, select Individual Search and then expand Image Label Filters. Then you can choose your viewpoint under Has at least one annotation with viewpoint (a logical OR for multi-select). Expand the Identity filter and you can enter the names/IDs of the whales to search. Then expand the Metadata and choose your username from the Assigned to user list to exclude any public submissions. When you submit your search, you’ll see an Export tab where you can download your files.
You can follow similar steps as above, but start with Encounter Search instead of Individual search. Select your viewpoint from Image Label Filters and then under Metadata filters, choose your username from the Assigned to user list.
As photos are added and IDs are assigned, the algorithms build on that knowledge. Both Hotspotter and PIE extract patterning and are tolerant to changes in patterning over time. As the new images come in with scarring and they’re positively matched to older images, the old image and the one with new scarring become part of the aggregated information.
Thanks for the follow up on this Anastasia. I tried following these steps (selected ‘right’ viewpoint, didn’t specify a specific ID, and selected our team user name Sheila Thornton from Assigned to User), but ended up with this error.
