If this is a bulk import report, send the spreadsheet to services@wildme.org with the email subject line matching your bug report
In which Wildbook did the issue occur? Internet of Turtles
What operating system were you using? (eg. MacOS 10.15.3) MacOS 15.6.1
What web browser were you using? (eg. Chrome 79) Chrome Version 140.0.7339.185 (Official Build) (arm64)
What is your role on the site? (admin, researcher, etc researcher
What happened? I uploaded a Bulk Import yesterday and sent to detection. The Import task status shows processing-detection, when I open the import it shows detection complete with 24 annotations (for 24 images), but when I open the first encounter there is no annotation Internet of Turtles
What did you expect to happen? I would expect the status to be consistent between the displays AND for annotations counts to be consistent.
What are some steps we could take to reproduce the issue? Bulk import 20250625 Bulk Import.xlsx
It doesn’t look like any of these encounters were annotated since the Class column is empty. I’ve resent this import to detection. I’ll check in on it later to verify it completed.
Hi Anastasia,
I believe some of the encounters were annotated, but not all of them. The status indicated that there had been 24 annotations. I did see that there were no ‘class’ indicators in the column, but that hasn’t always been populating for us.
I verified with the machine learning team that the reason annotations were limited is because this partially submerged angle wasn’t represented in the training data for Chelonia mydas, not because tehre’s an issue with the detection process. These will require manual annotations until retraining occurs.
ok, we’ll do the missing ones by hand - but the status suggests that they’re all annotated but they’re not - maybe some have 2 annotations? I’ll keep an eye out.
The complete status means that the detector reviewed all images and annotated all of the animals it found. In this instance, where the majority of the bulk import is made up viewpoints it’s not familiar with, what it really means is all images were checked and only a few turtles were found.
Retraining is the process the machine learning team goes through to review known individual data and manually retrain the detector for specific species. There is no set retraining schedule, but the more annotations your team has of identified green turtles to use for retraining, the better the likelihood of success for future automatic annotations.
