48-week year for GUI "Species by location" setting? #546
-
|
Hello, What is the 48-week year all about and how does one accurately implement that for the Windows GUI “Species by location > Week” setting? Thanks in advance! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 2 replies
-
|
The 48-week year is a constraint that eBird imposes, as they provide their occurrence data with 4 weeks monthly. So what we do is to take the current week with |
Beta Was this translation helpful? Give feedback.
-
|
Currently, the BirdNET-Apps use day of year for the calculation (so day of year is divided by 7.63 to get a week in range of [0..47]), as I thought this might improve the precision a bit, but afterall it shouldn't make a significant difference, because of the trained model. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for the info, I appreciate it. |
Beta Was this translation helpful? Give feedback.
The 48-week year is a constraint that eBird imposes, as they provide their occurrence data with 4 weeks monthly. So what we do is to take the current week with
date.isocalendar()[1], divide it by 52 and multiply it by 48. Yet, our geomodel is not very precise in that regard (which is intentional) so even if you provide a week that's a bit off (due to rounding or similar) it should still work fine.