Yep, here’s the one I always link to:
https://forum.inaturalist.org/t/annotate-dead-or-alive/17537
“if you use the time and place of original collection, it is Alive.”
My understanding is that the annotation is about the observation, not the image. I might be wrong, but I believe the CV model does not differentiate between alive and dead annotations, so providing that information to account for any “dulled colors” won’t have any impact. I’d be interested to know in more detail how the CV deals with observations for which there are numerous different “types of evidence” uploaded, whether it differentiates between them explicitly or the model is just built in a way that can accommodate such situations. I always think of a White-tailed Deer: iNat has images of full live deer, dead deer, shed antlers, deer prints, deer droppings, dead skulls, other bones, buck rubs on trees, etc. and the CV seems to nail them every time as “White-tailed Deer”, presumably without taking the dead vs. alive annotation into consideration at all. So I assume that by comparison “live bug” vs. “dead bug” isn’t that big of a deal for the CV to recognize, but I just don’t know.
2 Likes