Grouper Spotter Platform Error message when matching

Thanks for your patience and for the reminder to check in!

It looks like there was an issue where the match result did not get back to Wildbook and due to its age (jobs over 30 days old) it was purged from the machine learning server. You should be able to manually re-run the result to view matches.

The bad news is that this issue may have affected all of the encounters that were part of the same bulk import. So it’ll be up to you to decide it makes more sense to manually re-run match results on each encounter or delete the bulk import and re-import it again after removing the special characters from any file names.

Hi Anastasia,
Ideally, we wouldn’t want to have to remove and reupload the bulk upload because I believe the unapproved encounters that are currently in GrouperSpotter are from two different bulk uploads and some of the encounters from each of those bulk uploads have already been approved. So I’m not sure how we can easily remove and add those bulk uploads without duplicating some of the encounters.

When I rerun the matches or I remove the annotation (which is wrong in some cases) and add a new annotation to the encounter and then try to run a match this doesn’t appear to be working still. It spins for a long time (30 minutes), before it quits and leaves the message, attempting to fetch results.

So there is still an issue with these encounters. Unless I am misunderstanding what you mean by manually re-run the result to view matches.
I see you submitted a test encounter and were able to get a match but were you able to run a new match with the currently uploaded encounters?

Also is this going to be an issue later on when we want to rerun matches on some of these encounters that have already been approved? Or will it stop them from being able to match with future encounters? I guess I’m not quite clear on what the error was and whether it is fixed. We have used special characters in all previous bulk uploads.

Thanks for letting me know that re-running matches hasn’t worked and that redoing the imports isn’t an option for you.

@jason There seems to have been an evolution of matching issues here that weren’t fully resolved by January’s WBIA bug fix. Now matches vary from “Image Analysis has returned and no match was found” to “Gave up trying to obtain results”. Re-running matches or removing annotations to add new ones that start new matches also time out and don’t provide match results.

I can’t pinpoint when exactly the change occurred, but sometimes processes change as we continue to develop the product.

From our Bulk Import Cheat Sheet:

The docs say I can’t do XYZ in my spreadsheet, but I’ve done it before without a problem.
Wildbooks are always being improved as a result of testing and user feedback. Sometimes this means that we have to make changes to old processes to make way for new ones. We do our best to keep our docs updated to reflect the most up-to-date information on using Wildbooks. When in doubt, check the docs or post in the Community Forums for guidance.

If needed we can figure out how to remove and re-upload the encounters but it won’t be straight foward so if there is another option that would be ideal. I’m more concerned with previous encounters in these bulk uploads that have already been approved. Will they be able to be rematched in the future or do we need to redo them as well and if so how far back does the issue go?

For instance I just tried to rematch these previously approved encounters and the matching did not work. And when you look at the previous matches they all say “inspection image unavailable (likely outdated)”.

I worked with Jason on this one and because these encounters came from an import in October 2022, he just re-ran identification on the import and it appears to be showing matches again.

I thought re-running matches could only be done one-by-one, so I’m sorry about the scare with that. :sweat_smile:

When matches for an import are being weird, you can just re-run identification without having to re-do the whole import by clicking on “Send to identification” in the bulk import summary.

Thank you for the follow up. What about all of the other encounters on the page? The ones I tested that have been previously approved and in some cases matched I was not able to run a rematch for them. Do we need to rerun identification on all other previous bulk uploads if we wanted to rematch them.
If we don’t rerun the identification on these will they all be matched properly with new uploads?

Also should we change our naming convention for future bulk uploads and if so can you provide guidance on that.

Are they from different bulk imports? If so, you can go to the summary section of their bulk import page to restart identification.

You can do this anytime a match results aren’t displaying correctly from an imported set of encounters. If you’re not sure which bulk import an encounter came from, you can look at the Metadata section of an encounter and look for the link that says “Imported via [IMPORT ID]”. Clicking that will take you to the bulk import where you can send those encounters back through identification.

If they’ve already gone through identification previously, you shouldn’t have to. But if you want to re-run identification anyway just to see what new potential matches appear since their initial upload, I encourage it.

You’ll want to avoid special characters going forward. If it’s not A-Z, a-z, 0-9, space, or period, it’s a special character and may not play nice when uploaded. All of your file names should be distinct. We recommend using a file name convention that starts with the date so that all of your image names are unique.

Thank you for the follow up.

So for encounters that are in on the platform where match results are not currently displaying properly if we don’t restart identification they will still try to match with new uploads ok? But we can restart identification if we want to try to rematch them directly.

Or are these errors something we have to go back and check for regularly on all encounters to make sure all images are being used to match against new uploads?
From what I can see most if not all of the encounters are currently not displaying match results properly.

Yes :slight_smile:

No, not at all! Match results are just a snapshot in time of the matches that were available at the time the identification was first run. If matches appeared correctly when you first uploaded them but then don’t load correctly months later, you don’t need to re-run them unless you want to compare against any new photos that were submitted since your first upload. I’ve noticed this with my own older submissions in IoT.

It’s not a perfect experience, but part of the growing pains as we work to improve our products.

Hi Anastasia,
Sorry for the continued follow up just trying to make sure we have everything ok to continue to move forward.
When I look at our bulk imports

the last two don’t have Indentification listed in the last column and they don’t offer the option to “Send to Identification” Is there a reason for this?
Thank you

This is most likely due to the age of the import since those last two are from 2021 and the rest are from 2022.

But what does that mean. Those encounters can never be rematched? Is that correct? I thought we could always rematch encounters? Are they being matched with new encounters that are uploaded?

I was only referring to re-running matches on the images from those specific bulk imports. Those can always be re-done from the encounter page regardless of when it was imported.

You can. In the case of the 2021 bulk imports, you can re-run matches from the encounter page since the option is gone to do it from their bulk import pages.

GrouperSpotter will always review newly submitted encounters against all potential matches currently in its database. I think the part that may have caused confusion in my explanation earlier is in what it means to re-run matches on past encounters.

Let’s say an image of grouper 1 was uploaded in 2021. The matches for it that show up after uploading are based on a segment all of the images in the database that exist at that time. Now grouper 2 is uploaded in 2023. Its potential matches also include a segment of all of the images that exist in the database up until that point, including grouper 1.

Grouper 1’s matches from 2021 don’t have to be re-run unless the uploader wants to compare it against any new photos that were uploaded between then and now. If someone at some point determined that grouper 1 and grouper 2 were the same individual, that would be reflected in GrouperSpotter, even without re-running matches on that first encounter where grouper 1 appeared.

Let me know if that explanation helps or if i’ve just made it unnecessarily confusing. :sweat_smile:

Ok Great yes that all makes sense. Thank you for the follow up.

1 Like

Hi Anastasia,
Unfortunately more issues. The two bulk uploads that were still unapproved 2014 and 2015 are showing match errors again. They were ok a few days ok. I tried to send the 2014 one to identification but received an error for that as well…

Just an update the bulk upload from 10/25/22 seems to have successfully been sent to indentification but the one from 9/16/22 shows and error. And the questions still remains why do they both have match errors again?