New AI Computer Vision, First For Android
This is good news, it's always great to see things get better. You can learn about becoming a beta tester here. It will roll out later on for iPhone and the web.
One thing that would be great is to get the AI to stop suggesting things that don't really belong in our area. I did my part today as I added several species to this list-wiki of "stuff that needs curation" here. Since so many users are from eastern N. A., and that the expert user to casual ratio is so high for Fungi in a general sense and in particular for the East, it's like playing 'Whack-A-Mole' with these things. Hopefully, if we keep at it, next time I'll have to look a lot harder to find things that are clearly misidentified. I think Fungi is particularly challenging since our species concepts are so hazy. We don't even have pictures in all of the possible bins yet. Certainly not for any of the tough or large families. It is even harder to determine if something is in its range or not.
This is just another starting point for people who want to get involved. I tried to make it really easy to jump in and contribute by flagging all of my submissions. That way you can click on it and see my logic, see if anyone else has added their own, join in and add to the discussion or any combination of those things. Then just jump in and start taking things back to Genus or elsewhere if you don't feel confident with that. If you want to do yourself a huge favor have this up while you ID. When the little green shield icon is there in line with the species name that specifies that the ID was taken from an AI suggestion. That makes it even less likely that it will cause any arguments with the user or others who have chosen it. Use your judgment though, there is no reason to get upset or to upset anyone. Feel free to tag me if you run into issues or if you are having a discussion and are looking for help with an explanation. If you are local to N. A., ignore the ones without tags as they were unusual species or things I don't see as problems, here anyway. I think someone from Down Under must have added them.
Here is a search string to copy and paste to start in on this with for Russula sanguinea, then just go to the Russula sanguinea flags under curation and the copy and paste text is there in the comments. https://www.inaturalist.org/observations/identify?taxon_id=352308&&place_id=19,37,18,36,12,44,13,7590,6883,13336,7587,6853,9116,17,41,2,47,49,8,48,51,42,4,97454,39,33,7,31,29,20,32,35,24,38,28,25,36,19,26,45,30,43,23,21,27&per_page=200
Comentarios
Fungee, great point. I never understood why the AI will suggest species that are not actually in the area you are posting from. Seems like a simple fix. Perhaps there should be some way to over ride it with extra proof, as there is value in being able to post "alien" species to an area to show possible spread from one continent to another but its too easy to simply make the wrong choice on a common local species that has a look alike in another area of the world.
There is a way to add them as native or invasive, but it is complicated on another couple of levels. If you are interested @cooperj writes well about it on the tag I made for Russula sanguinea which I linked from the entry in the document I linked to in the post above. I still don't fully understand the CV and what it is choosing from. It does a great job at telling Russula from other mushrooms though. I just copied and pasted the same thing to about five hundred Russula observations in a row, I could have skipped looking at the pictures even!
Here are the locations for all states and linkable Canadian provinces in eastern N. A., you can refine your searches by adding &place_id=19,37,18,36,12,44,13,7590,6883,13336,7587,6853,9116,17,41,2,47,49,8,48,51,42,4,97454,39,33,7,31,29,20,32,35,24,38,28,25,36,19,26,45,30,43,23,21,27 to your IDENTIFY searches. This helps isolate the region where many of the problems I added need work on. I also like to add &per_page=200. I also make sure I check add all research grade observations as well as needs ID, and to check ANY under reviewed status. Then go to the page for the taxa and check to see if there is any guidance as to what to put things into, and if there is a blurb I've written. I'm working on them, some are more difficult to write than others, but mostly starting with the easy ones first, Russulas, and then in alphabetical order. I'm up to Cantharellus lateritius today, but I want to check out some more literature on that before I decide what we should do. It probably should be skipped for last as I don't think it is a huge issue since we may have it even. Maybe these might not be even blurb-able. It might be something worth going through to see if we can figure out which are which by sight alone. I think there are three, lateritius, flavolateritius and velutinus last I knew so maybe that is not an impossible task. Maybe it is best to be saved for last.
Comments might have to be reworded as I will copy and paste from the edited version of the comment which I don't think you have access to. That way I capture all the link wording so it is linked to the document using the format words.
Thank you for your clear explanations as well as information to help identify species. Also, thank you for correcting my misidentification of a russula which is most likely not found in North America. Your kind comments and links are greatly appreciate.
Thanks for doing this! You helped clean up an old misID of mine, and helped clear up a batch of bad IDs.
No problem, slowly but surely we will get there. I'm just trying to help everyone level up really quickly before my season starts and I am out in the field again.
Question. What if science proves the DNA is exact match and it was first described in Europe would it be a European name be the correct name to use everywhere in the world.
Well, that's the kicker, and the reason why I can't get the curators to just go along with my "evil" plan. Is that even provable that a Russula can't travel the world without a passport, or anything else I've taken up the mantle for? No, probably not even with DNA and you and me collecting. I just don't think it's going to be like the most popular Russula species wherever it gets to go to on some planted trees (which won't be the most popular trees wherever they happen to be planted) unless they are also invasive, or it jumps host as it appears to have done with Amanita phalloides in Cali. I don't know of anything offhand that is invasive that has that type of mycorrhizal partnership. And I don't know of any other invasive mycorrhizal mushroom that's jumped hosts. I would just rather the AI stop suggesting mushrooms that shouldn't be here, and start suggesting mushrooms that should. I'm also going to have to start ID'ing Russulas or some of these other things that the things I'm switching aren't, I'm assuming this too. I think we can do that, don't you, at least on some scale. If anyone can know them, it's us.
Re going back to Genus for say Russula or whatever, would it be good practice from someone to co-confirm at genus level and mark "as good as it can be," turning a record into Research Grade and taking it out of the main id/mis-id pool?
I'm always reluctant to say that something can't be identified from pictures since sometimes we find out that it is possible. Without knowing the extent of the speciation is problematic too. I understand that it might be a while though, and it truly probably won't be very possible for a lot of things. I would also hate to do that to someone who might sequence their finds or have it vouchered for later work where they might figure something out. I suspect the vast majority of the time it would be OK to do, especially the red ones.
Good advice, thanks!
I would LOVE it if you could manually add description features (viscid cap or smells like almonds) in such a way that the algorithm would take it into account in its suggestions! that's a real pipe dream tho.
Or habitat info!
Unfortunately I have a several someone mark in a way were it removed research grade when it was the first identified by its DNA and was used for inaturalist main page for that species. Now there photos are in place and some are not as good of a photo. I regret not reporting this to curators when it happened. Now I’m not sure what all ones this has happened to. A option to add nearby plants would be helpful and would be useful in filters box.
Feel free to tag me when that happens, I'll try to vote it back anyways. I know a way, but it would be a lot of work. Go to Mycoflora of Indiana page, find all of your linked observations, and compare the Mycomap name to the iNaturalist name. You can add the photos back yourself once it's named properly.
Añade un comentario