Is there a way to get it to stop showing me this dead rabbit?
Less problematic, but since I’m on the topic. I keep getting this terrible photo of a swamp rabbit. The desaturation makes it so you can’t see the color of the fur around the eyes which is important for IDing.
It looks like you already edited the taxon photos a few days ago. My guess, though I am not sure, is that the taxon photos used are those from when the model was trained. If so, they likely won’t change until after the next version of the model comes out.
That’s not what I changed. I didn’t remove any taxon photos for Eastern Cottontail. I added two and rearranged (including putting the old thumbnail back after someone had changed it to a juvenile). Who would have put a dead rabbit in the 12 photos anyway?
The one I removed for Swamp Rabbit is this one: https://www.inaturalist.org/photos/27674312
It’s the photo currently being used as the thumbnail for the genus and people keep shuffling the photos around to make it the thumbnail for the species as well. Having the same photo for both is confusing so I figured it would be less likely to be put back as the thumbnail if I removed it instead of just rearranging.
A photo of a dead rabbit could be a taxon photo for many reasons:
In this case, it looks like the Eastern cottontail has a bunch of subspecies with few observations, so that pic is probably a taxon photo for one of those due to the second reason above (but that’s just a guess). If you see it come up again, I’d recommend checking the photo and seeing which taxon/a it’s a photo for. You could then remove it. If it shows a key character, it might be best to leave it.
In regards to the desaturated swamp rabbit photo, I don’t think it’s necessarily a problem to have lower quality pics as some of the taxon photos in general, as they can show the species how it occurs under different conditions. Having a variety of views can help users understand the different ways a species may appear in photos. I have no idea if that’s why that photo was made a taxon photo, but if it is showing up a lot, that is an indication that a lot of people’s photos may be close visual matches to it.
Ah, hadn’t thought about subspecies photos. There are photos of rabbits lying down or standing up that show the belly. You are probably right about it being a gag. There was this one that has a very good shot of the hind feet color (diagnostic) but also shows the belly. https://www.inaturalist.org/photos/324600760 Someone removed it from the 12 but I put it back.
The desaturated one seems to come up when there are a lot of sticks in the photo. I think the CV is keying in on the background rather than the fur color. I thought that the desaturated one would be bad because Eastern cottontail has white around the eye. If someone had a photo of an Eastern cottontail and saw that photo of a swamp rabbit, they might select swamp rabbit because it matched. But really it is the other way around that is a problem, swamp rabbits being IDed as Eastern cottontails. So I think you are right about it being a good way to show that the color isn’t the only thing to look for.
Does CV re-train periodically on added photos? That is, if no photos of larvae are available for a butterfly species and I add some now, will these be reflected in say, six months time?
Should be reflected in less than 24 hrs in my experience.
The CV uses the entire image. Background will have an impact.
CV is updated about once a month. Due for the March update
https://www.inaturalist.org/blog/107012-new-computer-vision-model-with-over-100k-taxa February using January data
Needs 100 photos, 60 obs
(corrected - see tiwane’s comment below)
A new CV model generally comes out each month or so, sometimes a little longer. You can see some posts for new models on iNat’s Blog/journal as well as in the monthly forum updates.
The Model is about monthly. But the new taxon photo compare thing is 24 hrs about. If you change the taxon photos, should be reflected sonewhat soon.
Perhaps i misinterpreted the comment.
I know. That’s why I mentioned it.
That was the case in the beginning, years ago, but it no longer is. The page you linked to does not mention it because it’s no longer true.
A question: What if there are no larva photos? Google isn’t much help, and quite a lot of moths lack larva photos. If I wanted to, I could add one of mine. But I am not at all qualified to add a photo, both out of fear for a wrong id and because there is no one who is willing to confirm if my photo is of the right species. So is there anything people can do about it?
Make an obs with yours.
I already have, but no one has id’ed it so far.
Photos dont need to be RG to be added.
Ik but I meant no expert has identified them. I have no idea if the photo really is of (insert moth here), but it might be. Since there are no good photos, I have no way of knowing.
Oh. Thats just the way life is. Sometimes critters will only be identifiable at certain life stages. Or perhaps theres other reasons for this case?
I’m not sure. A lot of indian moth species are under-studied.
