In which Wildbook did the issue occur? ACW
What operating system were you using? Win 10
What web browser were you using? latest chrome
What is your role on the site? admin
What happened? Sightings uploaded in early August are showing up in the unapproved encounters search results as having been created 9/9/2020. Note that the bulk import logs show no BI’s done on 9/9/2020. Currently the system is showing 14 unapproved encounter records with that date on them. There are also a number of encounters showing as 9/8/2020 which should also be showing as an early Aug Create date.
What did you expect to happen? Created date should be Aug 4 2020 (or 5th for encounters generated by the system on these same imports). Or around that early Aug date - I did these E, F, G, H & I uploads (sightings by first letter of photographer name) over a span of a couple of days.
What are some steps we could take to reproduce the issue? Search for all unapproved records and the latest Created date in the results is 9/9/2020
These appear to be encounters created as a part of the same single import that is the subject of several other support tickets. Some of the images with a single animal have two nearly identical annotations for the same animal. This is treated as a second animal and a copy encounter was created. The original encounters created by the import have the date stamp of that date as expected, bu the duplicated were created on the same day we re processed these images to pick up missed annotations and thus display a later date.
We will look into this further, but it’s likely that there were improvements made to the detection model which was applied for the second run to pick up missed detections and a slightly better result was given that conflicted with the earlier run. Since there are only 14 that display this out of the import, if we don’t find a different cause than this I’ll just delete the duplicates.
@parham and I have investigated the 14 encounters referenced here. They are indeed products of re-running the one cheetah import from August that had missed some detections. Due to slight variations,
some extra annotations were returned for images that already had a detected animal and saved as another encounter. Since the encounters were created when the annotations were sent to detection the second time, the creation date on the encounter record reflects that.
These extra encounters were created as part of the manual re-run process on this import and should not ever occur during normal usage. We also now know to look out for this in the future if a manual re-run ever needs to be done again.
My recommendation is simply to delete these near duplicates, but I will wait for confirmation to do so.
Yes, I agree that deletion of these records makes sense. thanks!
Great! I’ve now removed the duplicate encounters, and only the original annotations and encounters should remain.