I’m a bit lot, I thought iNat was for sharing records of nature findings. If we should put more attention to AI training than to human identifying abilities, I’m not sure I want to keep be part of this.
4 Likes
iNat is for the human observers.
But us humans use the CV, and so we work together.
I do want iNat to have a better solution for habitat photos.
Why not? A new annotation for Habitat
?
4 Likes
I would encourage adding habitat photos, although I’m not an expert on anything. I would prefer seeing more about an organism’s habitat and preferences (which are important to any record) over the possibility of messing up the computer vision, which should not be blindly followed regardless.
3 Likes
We are in the early stages of using Computer Vision. The changes I have seen in the past five years lead me to believe that we will see more improvements.
In the future, Artificial Intelligence might be able to make use of historic habitat shots to add to our understanding of the natural world.Seeing a variety of associated organisms in a group in a certain spot at a certain time of year could tell us more than a close up photo of one flower. A number of habitat photos in one area can paint a picture. We just need to develop the tools to see it.
The iNaturalist Computer Vision appears to depend on the first photo in a series to make an ID.
That’s true; however, for training purposes, it pulls from a pool of photos not limited to the first from each observation.
1 Like
the population of the forum leans heavily towards identifiers and highly active users. this is why I think many make the mistake of thinking the CV should be prioritized – as we are the ones cleaning up when it makes mistakes.
But that isnt the case. the CV is (1) a tool, one that ultimately saves us a TON of work; (2) far less important then helping people appreciate nature and fostering curiosity.
with the current system, the most user-friendly and consistent answer is “make sure the organism is in every photo” AND “add as much info as you are willing / can”.
6 Likes
I think habitat photos can be extremely important adjunct to organism photos. Pity if the needs of the software override the integrity of the information
2 Likes
It can be very difficult if you managed one photo of the butterfly before it flew away or if, like me, you mainly record aquatic invertebrates. Getting the water beetle in the picture of the pond before you collect the water beetle is more or less impossible. I suppose after I have collected it, I could make sure the tube is in the corner of the photo. You wouldn’t be able to see the beetle but it would meet the requirement that the organism is in the habitat photo.
4 Likes
Exactly, try get the spider included in the habitat photo… I mean, technically it is there, but ususally not visible anyways.
I think habitat photos can be extremely useful and we need a better way to deal with them. I have an observation where I was the first observer of the whole genus, which is not well studied yet. So I felt a habitat photo would provide important information, and yes, I included it at the end of the observation for that reason.
3 Likes
I hope someone will make a feature request for a new Annotation.
Habitat.
Or if the subject lives buried in mud between mangrove roots.
2 Likes
How to make clear that the photo is one of the habitat of some mangrove-root-dwelling-butterfly, rather than the next photo (the roots of an unidentified mangrove plant) having been uploaded by mistake?
1 Like
I’m OK with habitat occasional habitat photos. It would really help if the observer commented something like, “photo #4 shows the habitat” so I don’t go searching through it for the bug – whether the bug is technically there but hardly noticeable or not. I don’t think every observation needs a habitat photo.
4 Likes
I’ll suggest that habitat photos could be uploaded to an external site (like Google Photos) and that a link to the photos is posted as a comment* if the observer thinks that showing the habitat is neccesary for identification of the species in question.
*it’s probably not a good idea to include links in observation notes, as the text is exported to GBIF if observation goes to RG
1 Like
Not a fan of using external sites to make my point on iNat…
3 Likes
I think everyone agrees that habitat photos are useful; the question is whether it’s best to include them in an iNat observation of some organism seen there, or post them separately and link to them. I see plenty of downsides to including them in the observation, as have been listed above, and really no downside to posting them as their own observation and just linking to them.
What I usually do is post the habitat shots as an observation of some large charismatic plant in the photo, like these:
https://www.inaturalist.org/observations/243459050
https://www.inaturalist.org/observations/243462351
https://www.inaturalist.org/observations/247013119
Then when I add observations of the little bugs and such that I find there, I can simply link to the “habitat observation” in each of those observations’ notes if I think it might be helpful. No external sites needed, no confusing photos uploaded as a species that isn’t present in the photo, and you get a few extra observations out of the day too.
4 Likes
That seems like a good solution, I think you’ve solved it! It’s going to be rare to have a habitat shot that doesn’t include visible evidence of at least one organism, and whether or not it’s identifiable to species doesn’t matter if its main purpose is as a habitat shot.
The only downsides are that it might be fiddly to insert the links, and I don’t know how easy this is to do via the mobile apps.
2 Likes
Do I gather you don’t ID plants? Distant shots that don’t include anything identifiable are one of my pet peeves…
If you marked them as casual, I wouldn’t mind, but there’s already more than enough in Needs ID that adding unidentifiable photos just for the sake of habitat shots is, well, a rather depressing idea.
4 Likes
And unscientific. The habitat is part of the info, should be part of chosen focus on an obs. Making a junk obs as a work around for - iNat doesn’t do habitat - is not a solution.
Plant growing in water. Is it a pond, or a stream. Or on the beach. Does it matter. For an ID.
3 Likes
We could use observation fields for this. A habitat one already exists. Given all the possible habitats an organism could be in, an annotation for this might not be feasible.
Observation fields also have the advantage that they can be added directly on the upload page.
1 Like