Can you describe what the issue is you’re experiencing? Some match results display nothing, others display that they are fetching results, while some encounters have no match results.
I believe this may be in part because of the size of my bulk import, but it has been a few days at this point and I’m wondering if these are the results after the system has completed identification, or if the system is still working through this bulk import? How will I know for sure if identification is completed?
Can you provide steps on how to reproduce what you’re experiencing? Bulk import and examples are linked above. I’m attempting to look at match results for encounters in bulk import.
If this is a bulk import report, send the spreadsheet to services@wildme.org with the email subject line matching your bug report
Hi, @BTran!
I find myself wondering whether this has to do with the half-deleted bulk import. Maybe the system barfed a little bit when it found assets that it was trying to identify that were missing because of deletion.
I’ll plan on looking into this in more detail on Thursday/Friday of this week.
Thanks,
Mark
I want to also say that if it’s possible, we’d be okay having everything removed and starting off with a clean slate on the website, if it makes things easier.
We have the spreadsheets and all the photos saved up for bulk importing, so we can reupload everything again in smaller batches if it comes down to that.
I think a workaround for you in the meantime would be just doing a bunch of smaller bulk imports. The duplicated encounters from the current bulk import will have different uuids, so deleting them later won’t affect ones that you upload now.
Another issue with some of the smaller bulk imports I wanted to bring up.
The first few bulk imports seemed to work alright, but then a few of the more recent ones would confirm that images were sent to identification but then would not actually do anything when checked on after a few hours (ends up asking if we want to send to identification again): https://www.grouperspotter.org/imports.jsp?taskId=b8a72f09-975f-4290-8ee6-06a1b2542ce7
Before and after images of the same bulk import (linked above):
Hi, @BTran !
Our machine learning expert reported some slow behavior on grouper over the weekend and into today. I’ll look at this asap. There’s been a lot of activity on community lately, so the backlog has been growing. But I have not forgotten you, and you’re definitely on the queue!
-Mark
Would you be comfortable of giving an ETA when you’ll be able to look into these issues on grouperspotter? Later next week, within the month, etc…?
Hi, @BTran !
We are noticing similar issues in a few other wildbooks.
I can’t give you a solid timeline because we don’t yet fully understand the nature of the problem. I certainly think that, “within the month” is realistic for a full fix, but we can definitely try to work around it!
Are you able to delete the new, smaller bulk imports and retry those that failed?
-Mark
Hi, @BTran !
We killed the jobs on grouperspotter and re-started them.
It’s got ~41,000 jobs in its queue (the samples get compared against each other).
Can I request that you don’t submit any new bulk imports in next, say 4-5 days?
In the future, if y’all anticipate having huge bulk imports like this, limiting the ID by locationID might be super useful for speeding ID up.
I’m currently deleting the large bulk import right now, but afterwards, can we continue uploading smaller bulk imports to the site, at this point in time?
Also, I’ve been reading through some of the documentation on image analysis and the matching system used by WildMe, but I’m not too familiar in how everything works just yet.
Are matches limited to the images provided within the bulk import? Do they look at previous bulk imports as well? If we wanted to compare all images to all other images, should we upload everything first and then start the matching process for each bulk import?
Hi, @BTran !
I think that it might be time to check on these identifications, now that the big deletion has completed.
Regarding your question,
Are matches limited to the images provided within the bulk import? Do they look at previous bulk imports as well? If we wanted to compare all images to all other images, should we upload everything first and then start the matching process for each bulk import?
Matching happens independently of bulk import. Bulk import is just a way to get lot so encounters into the system more quickly. So, yes, encounters from previous bulk imports can be included as potential match candidates, provided they have the same viewpoint (e.g., left or right side of the fish).
