Reporting errors on pilot whale algorithm

Hi,
On Flukebook for this bluk import for short finned pilot whales: https://www.flukebook.org/import.jsp?taskId=1da4e3a1-58e2-4344-ba78-525ae2ba5a98

I noticed when the detection consider the entire image as the annotation, identification even completed will be stuck in “attempting to fetch results” and I never can access the matching table. However if I use another annotation the previous “attempting to fetch results” annotation can correctly be proposed as a candidate.
I have to manually remove and recreate all annotation that are detected as the entire image.

Additionally, the pilot whale algorithm never propose corresping images of opposite sides of the dorsal fin. We have to manually be extra careful to not miss matches with opposite sides and not create duplicates.

I hope this can be fixed.

Best regards

Can you describe what the issue is you’re experiencing?

Can you provide steps on how to reproduce what you’re experiencing?

If this is a bulk import report, send the spreadsheet to services@wildme.org with the email subject line matching your bug report

Hi @lsimon

I looked at a few of the encounters in this import and all of the images I saw are of closely cropped dorsal fins. It’s not unusual for the detector to annotate the entire image in that instance. It doesn’t sound like the missing initial match results were due to any issues with the annotations, but likely just an issue when it went through identification that was corrected when the match was re-sent. If this is an issue you’re seeing with all of your pilot whale uploads, let me know because that points to some other issue.

That’s by design. Match candidates are determined by the taxonomy (only match against same species), viewpoint (such as matching left sides with other corresponding left sides like left front or left back, etc.), and location ID. If the wrong viewpoint is displayed for a match candidate, it could have been mislabeled in the encounter.

Thank you for the feedback, I’ll tell you if the annotation issue happens again.

What do you suggest that I do to ensure that the left side of a fin can be propose as a match with its right side for short finned pilot whales ? For other dolphin species it is usually not an issue.

There isn’t anything you need to do because that’s not how matching is meant to work in Wildbooks. Similar viewpoints are only supposed to be matched with each other.

Ok I didn’t know that thanks. I thought so because working on Fraser’s dolphin, Right and Left side are usually compared and can be match on Flukebook. I’ll just make sure to compare them visually to avoid creating duplicate individuals then.

Thanks for your answer

1 Like