3 ms·

That seems like one of the problems that can be solved with technology, i.e. by using - machine learning could be used to visually identify fungi - Portable D

That seems like one of the problems that can be solved with technology, i.e. by using - machine learning could be used to visually identify fungi - Portable DNA sequencing devices i.e. nanopore? https://nanoporetech.com/products/smidgion https://nanoporetech.com/products/smidgion

_whiteCaps_ 4y ago
No, machine learning to identify the fungi is the problem: > So he did what he thought was the responsible thing: he used the plant identification app on his phone to determine the species of mushroom. I consider myself a beginner at foraging wild mushrooms, and even I can tell that the apps are laughably bad at identifying them.
- sva_ 4y ago
  I think they can give you a first idea of what you should be searching for, if you find a specimen you have no name for. From there on you'll have to figure out what is similar, edible, and grows in your area; and in particular, how the edible ones are different from inedible/poisonous ones. It is usually just a few. I can't fathom eating some wild fungi without doing this kind of due diligence, blindly trusting an app.
hutzlibu 4y ago
"machine learning could be used to visually identify fungi" Even if the recognition rate would be way better - NOPE. There are mushrooms that look really identical - but one is deadly poisenous and the other good to eat. Some mushrooms you can only identify by slicing them up, smelling, watch color change, look for certain small details under the top - and where they grow. With current state of ML, you can get only a very rough classification and I doubt that changes anytime soon. Portable DNA sequencing sounds better, but I am sceptical, if the tech is mature enough. (If it is, sign me up, but it sounds expensive, they do not even have a price tag)
pfdietz 4y ago
This sounds like a problem that can be solved by - not eating wild mushrooms
pazimzadeh 4y ago
*eventually be solved Also I think you need to perform some kind of DNA extraction on the fungal tissue before using the nanopore?
rnhmjoj 4y ago
No respectable micologist would attempt identifying a mushroom based on a picture alone. It's just not possible: there isn't a 1:1 correspondence between species and (macroscopic) appearance, not for every species or even the majority of them. To be relatively sure on an identification you can combine a good knowledge of the location where the mushroom was found, smell and even taste, but a microscopy (observation of the size, shape and color of the spores, characteristics of the sporophore, hymenium, etc.) is necessary to be certain. In fact, the taxonomy was entirely based on it before genetic analysis started to take over. You could probably do machine learning and reach a good accuracy based on microscopy, but certainly not with a blurry picture taken from a smartphone.

_whiteCaps_ 4y ago

No, machine learning to identify the fungi is the problem: > So he did what he thought was the responsible thing: he used the plant identification app on his phone to determine the species of mushroom. I consider myself a beginner at foraging wild mushrooms, and even I can tell that the apps are laughably bad at identifying them.

sva_ 4y ago
I think they can give you a first idea of what you should be searching for, if you find a specimen you have no name for. From there on you'll have to figure out what is similar, edible, and grows in your area; and in particular, how the edible ones are different from inedible/poisonous ones. It is usually just a few. I can't fathom eating some wild fungi without doing this kind of due diligence, blindly trusting an app.

hutzlibu 4y ago

"machine learning could be used to visually identify fungi" Even if the recognition rate would be way better - NOPE. There are mushrooms that look really identical - but one is deadly poisenous and the other good to eat. Some mushrooms you can only identify by slicing them up, smelling, watch color change, look for certain small details under the top - and where they grow. With current state of ML, you can get only a very rough classification and I doubt that changes anytime soon. Portable DNA sequencing sounds better, but I am sceptical, if the tech is mature enough. (If it is, sign me up, but it sounds expensive, they do not even have a price tag)

pfdietz 4y ago

This sounds like a problem that can be solved by - not eating wild mushrooms

pazimzadeh 4y ago

*eventually be solved Also I think you need to perform some kind of DNA extraction on the fungal tissue before using the nanopore?

rnhmjoj 4y ago

No respectable micologist would attempt identifying a mushroom based on a picture alone. It's just not possible: there isn't a 1:1 correspondence between species and (macroscopic) appearance, not for every species or even the majority of them. To be relatively sure on an identification you can combine a good knowledge of the location where the mushroom was found, smell and even taste, but a microscopy (observation of the size, shape and color of the spores, characteristics of the sporophore, hymenium, etc.) is necessary to be certain. In fact, the taxonomy was entirely based on it before genetic analysis started to take over. You could probably do machine learning and reach a good accuracy based on microscopy, but certainly not with a blurry picture taken from a smartphone.