Id say it’s better to have habitat shots than not because in many cases they are the defining factor in identifying certain species. I do think a workaround in terms of CV is needed rather than doing away with the photos entirely.
5 Likes
I agree. If it’s a choice between coddling the CV or showing features of the habitat, I vote for the habitat - more useful and more interesting.
4 Likes
I would describe the habitat in the comments section if it is important to identification. Sometimes a photo of habitat does not tell the whole story. If soil type, temperature, if wetland (what type) etc. is important, the photo would not necessarily tell you this.
2 Likes
I don’t see why such photos would be any more of a problem for the CV than habitat photos where the organism is not present (as I noted, I suspect on the whole these are not a major factor in the CV learning; if the plant is distinctive enough to be identified from a tiny fragment in a photo, there is a good chance that there are plenty of observations for training the CV, most of which are likely to have photos where the organism of interest does feature prominently).
I do think that users should be discouraged from simply duplicating observations in such cases instead of cropping/marking the photo to point out the organism of interest.
But this has nothing to do with the CV and everything to do with consideration for other people who might be looking at one’s observations. Even if the observer notes what they are interested in, it is easy to overlook a textual note and difficult to overcome one’s instinctive reaction to what appears to be the focus of the photo. There are a few users whose observations I regularly skip simply because I find it tiring to play “where’s Waldo” all the time.
2 Likes
I do the same. It’s not difficult to include the organism in the habitat photo, and can be useful for establishing the size/scale of the organism along with the environment it is found in. If the CV can handle shots of different plant parts (leaves/flowers/fruits), it should be able to handle different views of the organism as well I hope? (i.e., close-ups, full-size, silhouettes/outlines)
4 Likes
I think habitat is important and have been asked to include it on some obs.
The order of photos from close up through “parts” with close habitat last, in my opinion should work without messing with CV.
There can be a written description in the notes, as well. I find these can be overlooked.
2 Likes
As others have noted, the best solution if a habitat photo is desired/needed is to take one with the organism in it, even if it isn’t the focus or is small - this is totally fine. There are many reasons why including habitat photos without the organism in them can cause issues, some of which are mentioned here, but there are also extensive discussions elsewhere on the forum. They include issues with CV training, confusion of IDers (who may not know a photo is a habitat shot, may waste time trying to find the focal organism or ID for an organism that isn’t the focus but is in the habitat shot, might use the single subject DQA, etc.), issues with photos showing up for the taxon on iNat that don’t have that taxon in them, photos being exported to GBIF that don’t have the taxon, etc. Habitat photos (without the focal organism) also just break the fundamental assumption of an iNat observation that an observation’s media contain the organism they are for - this can be a problem for any downstream users who use iNat observsations/media in ways we might not necessarily predict. Until/if such time as photos can be designated as not containing the focal organism in some way (which would solve a lot of the issues discussed above), @jdmore’s solution is an easy way to address this that “makes everyone happy.”
8 Likes
However, it will tell a lot more of the story than what somebody that doesn’t know how to answer any of those questions (or which questions are relevant) will do. It’s very easily providable context, without a required minimum spec of base knowledge from the user.
4 Likes
I agree. With plants and fungi so many times people ask me what something was “growing near” or “on” and I’ve included some habitat shots for that very reason (I was not aware of how it might impact AI training until reading this thread…and I can think of a few where I need to go and remove them, like my recent owl pellet obs…which I added because I wasn’t sure if it was a pellet or scat and wanted to show where an owl has been seen perched). I would love the option to include a photo in a comment below to show the habitat. I think it is important in many cases for plants and fungi.
Someone in this thread suggested asking the user to mark photos to exclude from the CV, but I think that’s asking too much…I believe many will not do it (so many people already do not take time to add any initial ID such as plant or animal and just upload as unknown…I imagine they would not take any extra step to mark something). Would they take an extra step to put a habitat photo in the comments? Probably not, but this option would allow those of us who think they are important to put them somewhere.
It is possible to embed images in a comment: <a href=“address of web page where image is located”> <img src=“image address” width=“250px”> </a>
1 Like
A lot of the “habitat” photos that I have seen don’t usually give much information that would contribute to the ID. If you want more information, I think it would be more useful to make a number of separate observations within the habitat and note it in the comments.
I agree, for the very reasons you give (though possibly the sort of people who are more likely to include habitat photos are also more likely to carefully mark them?). But if it could be something others could do, people who care and notice could ‘fix’ it.
1 Like
All good points!
In also wondering if this works: I take a picture in a plane or a mountain range and upload it on iNaturalist as an observation of a plant that is found in that range?
If the something is what ? A tree big enough to see on satellite view ?
No, but rather a tiny plant that is known to be present in the range but not at all visible in the photo.
If you did - you would probably attract a DQA
‘no evidence of organism’ and it would be Casual.
Known to be present - is not enough for an iNat obs - you should see it for yourself.
9 Likes
You’re right, maybe many observers (the majority?) would never add a habitat photo, and maybe in many cases it wouldn’t add significantly to the information in the main photos anyway, but it seems a pity to “level down” on the quality (always a contentious word, I know) of the observation when it can be avoided. If additional information (such as a habitat photo) is available, I personally believe those users willing to put in the extra time and effort should be encouraged and facilitated to include it in the observation. Obviously in such a way that it doesn’t “warp” the CV and/or cause any other technical panic attack 
.
Yes indeed, and I have actually used this work-round a few times. But it’s a bit off-putting unless you have a particular type of technical know-how and you also need somewhere stable to host the photo. I use my site, but it’s not something you can probably expect the majority of observers to want to/be able to do.
4 Likes
What is CV ???
As of dec. 25, 2024, I can’t find any coherent definition in the wiktionnary for this ?
iNaturalist lingo for Computer Vision, see:
https://help.inaturalist.org/en/support/solutions/articles/151000170364-what-is-computer-vision-
Also sometimes referred to (incorrectly) as AI.
6 Likes