App having trouble IDing slimy stuff

I am seeing a lot of what is apparently Sap Yeast (Cryptococcus macerans), but I have never gotten it identified through the app. Instead I get Stinkhorn, Lobster Mushroom or some higher-level taxa such as Basidiomycetes or Agaricomycetes.
I am getting lots of angles and magnifications, so I don’t think doing more of that will help.

1 Like

hey @jimharkness, welcome to the forum!

the ID suggestions are generated by an AI which is based on iNaturalist observations. Due to its nature, in order to learn to suggest a taxon, it needs to be fed many photos of correctly identified observations. Some taxa are not available since there aren’t enough photos of confirmed specimens to feed the AI. This is common when a species is hard to identify or requires very specific types of photos for ID, like macro shots of particular parts. Other times, the suggestions simply are inaccurate because of false positive IDs.
the best thing you can do to help is to observe and identify the species in which you are interested, contributing more data and photos to the AI. maybe the next time the AI is updated, it will be able to ID your species of interest :)

6 Likes

Welcome, @jimharkness! Those kinds of less commonly observed organisms really need a human eye. I see you haven’t posted any observations to iNaturalist yet, so you should do that to get community input. If you only want computer vision suggestions, check out Seek by iNaturalist. Hope that helps!

2 Likes

When I come across obs like this one, I generally identify it as Fusicolla merismoides and mention that it is also known as ‘slime flux’, which is made out of a variety of fungal and bacterial organisms.
I first learnt about it through the ‘slime mold identification & appreciation’ facebook group where slime fluxes are regularly posted.
Someone there with more expertise listed some of those organisms (more than 20, including C. macerans) :-)

I wouldn’t be surprised if it also gets IDed as some of those other taxa, which would ‘dilute’ the training sample for the computer vision algorithm.
I guess it would be nice to have the possibility to identify some obs at such composite level, in this case ‘slime flux’, since it is made of several taxa, but is still somehow recognizable as an entity.

2 Likes

Hi @jimharkness The most recent computer vision training run had a threshold of “at least 100 verifiable observations and at least 50 with a community ID [as of September 29, 2019] to be included in training”.

There were 63 observations added to iNat before that date that are currently identified as Cryptococcus macerans, so iNat isn’t suggesting that species because it’s not available in the computer vision model. The good news is that there are now 93 observations with that ID, so we’re not a long way from this species being in scope for a future training run. If you’re knowledgable about fungus identification you can help out by spending a little time in the Identify tool looking for likely candidates to ID.

4 Likes

Hi Carrie! I was using Seek when I had this issue: I did not realize there were two different apps! Will try again in iNat, which I just downloaded. #totalbeginner

3 Likes

Thanks for this. Tragically, I am not a fungus experts and was just trying to use the Seek app to get an ID. I will try to get more pics and share on iNat.

1 Like

Thanks. My challenge is that my deepest knowledge is birds, which are hard to photograph. I know nothing about plants and fungi, and was just hoping to get a computer ID. I will follow Carrie’s suggestion and share photos of unidentifiable species with the community. Maybe the fungi experts can help make IDs and add to the pool.

4 Likes

You’re right that extra photos won’t yet help get an automatic ID suggestion for your observations, because computer vision doesn’t yet know about Cryptococcus macerans. But these extra images will be a big help once computer vision does start to work on identifying this species.

Also, my understanding is that the same computer vision modeling work is used by both Seek and iNat (although updates may happen at different times).

As for helping with IDs, you don’t need to be an expert to help out. It’s fine just to work on a particular type of organism or particular area that’s familiar to you. So long as you take care to make the best ID you can, that’s a big help. If you add comments (“I think this is X because it has longer horns than Y”) that really helps too.

2 Likes

Just one more thing… I think iNat will probably be more satisfying for you than using Seek alone. The big advantage is that other iNat users can add their suggestions about IDs for what you’re seeing. It’s quite rewarding to have others confirm the IDs that you figure out. And when people say “Actually, it looks like it’s a xxxxx instead” (as with every spider observation I upload) it’s a great way to learn.

6 Likes

One thing that makes birds easier on iNat is that there are so many people who avidly ID them. I’ve uploaded some very bad bird photos and had multiple IDs, all agreeing, within minutes. Bird photography does take some practice, and a zoom lens (or point-and-shoot with a built in zoom) certainly helps, but I think you’d be surprised how forgiving iNat is of less than perfect bird photos.

3 Likes