In which Wildbook did the issue occur?
Amphibian & Reptile
What operating system were you using? (eg. MacOS 10.15.3)
Win 10
What web browser were you using? (eg. Chrome 79)
Firefox 121.0.1
What is your role on the site? (admin, researcher, etc)
researcher
What happened?
Some matching results are categorized as “pending”. When opening the matching results page, only the left image is displayed, but not the right one.
What did you expect to happen?
Both images to be shown so a match could be confirmed (or not).
Was this Encounter deleted? When I try to use the taskID in the URL to find the encounter, it shows up as no images or annotations associated with this match result.
Thanks for re-running the first example. The second one seems to have been matched in the meantime by another team member of ours.
I will re-run similar cases when they occur.
Hi all,
I was working in the same project (Amphibian and Reptile Wildbook | Login) and I had two encounters with the same problem (Identification Status: Pending and under Match Results a score is given but no match candidate):
It’s hard to say without having the link to the original match result (the URL to the match changes each time a new match is run). Deleting and redrawing annotations on the match candidate images can impact their appearance in previous matches since matches are compared by their annotations and not the full image itself.
Here is a similar case where I have not re-run the match to leave the URL untouched for you to review. Is there a log somewhere of previous match results? That would be extremely helpful for these cases.
Interestingly, I noticed with this encounter that if you choose “Botanical Garden Bielefeld Adults” under “Project Selection” (the project it is assigned to), we don’t get a match candidate nor score but when we choose “None Selected”, we will get the score and nothing else!
I verified that this was a result of the match candidate that was available when it was first run through ID in Nov 2023 had its annotation deleted. This is why you see a match score with no corresponding match candidate image. Starting a new match will show you an updated list of match candidates with current annotations.
Each attempt at running a match creates a unique task ID (the part of the match page URL that comes after iaResults.jsp?taskId=. When you click on “match results” from the Encounter page, it will only take you to the latest task ID/match attempt. Unless you happened to save the link to the previous result or still have it open in another tab, there isn’t a way on the user side to retrieve it.
A match score with no candidate image sounds like an annotation was deleted in between when ID was run and the Encounter was reviewed. Try re-running the match from the Encounter to see the most up-to-date results. It’s best to review ID results as soon as you can to prevent issues with the match candidate data changing before you’ve had a chance to review it.
Thank you for your review and clarification. The dilemma we are facing then is that quite a few annotations are incomplete (e.g. missing a portion of the salamander head of body). If we delete these annotations and manually draw new ones, we risk losing existing matches (because they do not always show up again with a new annotation and match attempt). If we don’t re-do the incomplete annotations, we risk having less matches in future.
Is there a chance to implement creating a logfile for each encounter that logs all match results, so they can be reviewed even is the annotation is changed or/and a new match is started? (I can also post this under “feature requests” if necessary.)
Another helpful alternative that might technically be easier to implement would be to show on each encounter page which matches with other encounters are currently proposed, but not yet reviewed (“pending”). That way, before re-drawing an annotation, these proposed matches could be seen on the encounter page and - if needed - merged before re-drawing the annotation and starting a new match.
That log sounds helpful.
My guess is, though, that the match candidates “we lost” after drawing the new annotation (and are being only represented by the score anymore), are incorrect anyways as they are based on an imperfect (or not so good) annotation?
I suspect it’s unlikely for data storage reasons. Also, if there are plans to bring MiewID to ARW in the future, that would likely resolve these issues as it outperforms Hotspotter in the Wildbooks we’ve introduced it to. You’re welcome to submit feature request for it, though.
Edited to add: Or you could just leave both annotations there without deleting the incorrect one. ARW will clone the encounter and you can view the matches for each annotation on its respective encounter page. It could make data management more complicated because they aren’t two different individuals, but two encounters of the same individual annotated twice in the same image. However, if you’re determined to have access to matches with both the automated annotation and the manually drawn one, that’s the easiest way to go about it.
That’s the idea. This mostly comes down to data management workflow. If you bulk import Encounters, wait to send them to identification until there’s time for matches to be reviewed. This saves researchers the frustration of making decisions with outdated data that may have been changed since the match was first run and reduces strain on the servers since we won’t be re-running the same jobs twice.
@David.Kupitz made a good suggestion yesterday that we have started following in our workflow which is to
check all encounters for Match Cadidates and merge/mark new depending on whether a correct match candidate was suggested even if the annotation was imperfect
note down which encounters have annotations that should be manually re-drawn
after matching has been done for all encounters, re-draw annotations and start new matches for the encounters noted under 2)
I hope this way we can avoid issues with annotations having been removed and other issues due to “the lag” between identification and checking match results.
