What is the entire URL out of the browser, exactly where the error occurred?
Proposed match - source encounter that doesn’t have a viewpoint assigned: Wildbook for Carnivores | Login
Match results page where above encounter is proposed match #3:Wildbook for Carnivores
Can you describe what the issue is you’re experiencing?
I ran a match for an encounter record and in reviewing the proposed matches, I saw that the 3rd one is of a leopard with the opposing viewpoint. When I went to check the viewpoint assigned, I can’t see any sign of a viewpoint having been assigned.
There’s no viewpoint showing when I hover over the annotation in the encounter that is the source for match #2, which is the reason it’s not being excluded from the match results based on it being the opposite viewpoint from the target encounter.
For reference - the 2 examples above came from a spot check of 3 encounters’ match results in an upload batch from yesterday.
We believe this import was submitted right before a fixing change was made to the leopard detection config. Do you have access to the excel file for this import? Removing and resubmitting it will most likely fix the issue. I’ve taking a few samples of images with missing viewpoints and had successful full detection results with viewpoint come from them with the current config.
Hi @colin - these batches were submitted in April. Much work has been done with these images - at least one merge between 2 individuals and matches with other encounters. What happens to all of that work if I delete and re-submit these batches? Am I correct in assuming all that work will be lost? If so, then that’s not an option.
Is there no way this can be fixed on the back end? Have the detector run on these images and any other leopard batches that were uploaded prior to the fix, in order to get the viewpoint assigned?
If not, that means we have hundreds of leopard images without viewpoints??
After conferring with Jason H, we think there is a possible solution. We can remove all annotations without viewpoint and resend just that set to detection for new ones with the corrected config.
This is an option with leopards because they are solitary, images with more than one animal have the potential for encounter duplication.
We can proceed with trying this per your approval.
Hi @colin - if that would retain all of the matching etc already done in ACW but apply viewpoints where they don’t exist today, with the only downside being that where there are images with multiple leopards we may get some duplicates, I think that’s a good solution. As you say, they tend to be mostly solitary so the # of images that have multiple leopards in them will be fairly small, making a cleanup of any duplicates fairly easy.
Please go ahead with this and thanks for figuring out a solution! Phew!!!
Hi @jason - that’s great, thanks so much! I’ve only checked a couple but I noticed that the match results previously run are no longer available - is that a consequence of this process?
Yes, those match results were associated with the old annotation. The updated annotation with viewpoint is a new annotation requiring a new matching Task.
Doing some spot checking, I found some media assets without annotations. So I added annotations manually, and noticed that even though I added the viewpoint, as required, it doesn’t show when I hover over the newly created annotation in the encounter record’s gallery. I assume this is just a ‘cosmetic’ issue and that the system does know what viewpoint I assigned and thus matches run will only be run against that same viewpoint, but I thought I should mention it.