Workflow for correcting bad dateTimeOriginals

[nathanielrindlaub]

Not infrequently, dateTimeOriginals that camera traps bake into their images can be inaccurate, either because the user input the incorrect date into the camera settings or because the camera reset itself to some default start date. It would be awesome to provide functionality and a workflow through Animl to correct this.

Des recently used Animl to ID images with bad dates, then exported them and corrected them in R using the workflow described below. It may be worth replicating this or something like it:

Thanks Natty, and you bet! I spent a lot of time with Tim deciding how to fix these dates in R, course the final method is still a little brute force. If you have any suggestions I'm happy to try something different!
Finding the reverted cameras was easy with sorting by date - some of my Bushnells switched to their default manufacturing date, which was 1/1/2012, 1/1/2018 or 1/1/2019 depending.
I found the bounds of the incorrect dates by checking the 'person' labels, since that's usually when we'd correct the camera date/time.
Once I had the bounds I went into the .csv in R and specified the rows to be fixed, then found the # of seconds between the dateTimeOriginal and my correct datetime. I added that value to the 'bad' rows. Using seconds to add time helped account for daylight savings, leap days, etc.
Then I replaced the incorrect rows with the corrected datetimes! One camera (RS3) reverted twice, and we were able to use 'date added' to parse out different photos with the same 'date created'.