fieldherping.eu

Photo based species recognition softwares are rapidly being developped nowadays. Think of Plantnet for plants, ObsIdentify for inverts in the benelux,... and today I found this, until now, browser based photo recognition app for American herps (associated with the herpmapper team). It is still a demo version but performs surprisingly well! Here is their gofundme page: https://www.gofundme.com/computer-visio ... h-hardware. Go check it out!

http://vision.herpmapper.org/

Pablo

Not bad...BUT, do you think such software could ever replace guide books and human expertise? It seems unlikely to me for the simple reason that it can't think. Sometimes there are too many variables for a computer to deal with, e.g. I tried with an image of Graptemys barbouri juvenile and it went wrong by far, i.e. the correct species was not even on the list of suggestions...I know it's a demo version but I don't see how such a software could be 100% accurate, or at least more accurate than a human.

I also think it gives you far too many suggestions. If you're trying to identify a species and it gives you a list of ten (spanning across two-three families) for which the percentage is not conclusive that's not much help. In fact one would probably know more(at least the family) than this even before one tries to ID the species with the software.

No way that these apps could replace field guides in any way. There will always be gaps because some species are highly variable in appearance or look different when they are juveniles or even before shedding. But I don't really think that the aim of these kind of apps should be to get experts/advanced field herpers to use them. But it will help interested people to identify those commonly encountered species, those are not the people that already purchased field guides (the kind of people that post their images on facebook species identification pages). I don't think that herpetology as a science benefits from this, it is just cool that this is possible . Let's just dream of affordable sequencers in the near future that come with a blasting app on your smartphone...

I agree Pablo!

Pablo Deschepper wrote:Let's just dream of affordable sequencers in the near future that come with a blasting app on your smartphone...

Not sure if it's a good dream or a nightmare though

Ilian Velikov wrote:I agree Pablo!

Pablo Deschepper wrote:Let's just dream of affordable sequencers in the near future that come with a blasting app on your smartphone...

Not sure if it's a good dream or a nightmare though

Hah why nightmare? Do you fear eugenics among humans and discrimination by insurance companies based on DNA?

Michal Szkudlarek wrote:Ilian Velikov wrote:
I agree Pablo!

Pablo Deschepper wrote:
Let's just dream of affordable sequencers in the near future that come with a blasting app on your smartphone...

Not sure if it's a good dream or a nightmare though

Hah why nightmare? Do you fear eugenics among humans and discrimination by insurance companies based on DNA?

I fear many things. Humans don't exactly have a clean history record when it comes to possessing power.

Ilian Velikov wrote:

Michal Szkudlarek wrote:Ilian Velikov wrote:
I agree Pablo!

Pablo Deschepper wrote:
Let's just dream of affordable sequencers in the near future that come with a blasting app on your smartphone...

Not sure if it's a good dream or a nightmare though

Hah why nightmare? Do you fear eugenics among humans and discrimination by insurance companies based on DNA?

I fear many things. Humans don't exactly have a clean history record when it comes to possessing power.

Ancestral sin?

Just some thoughts about the reply of Ilian about the variabilities, diferences due to the age, etc. A herpetologist relies on a number of parameters to identify a specie. The more experiences he gets the more accurate he will be. Always 100% reliability? I'm not sure. I remember a discussion a few months ago on this forum about the identification of "brown" frog. Was it a dalmatina, temporaria,...? Even the best specialist on this forum would n't agree immediately.
That saying, knowing a lit bit of artificial intelligence, I am quite convinced that "an app" would be able to identify a specie with the same confiidence as a human IF and only IF it is fed with the right parameters ( and that can be quite a lot). There are a lot of mathilatical tools like Support Vectors Machines, Hidden Markov models, Neural networks which can handle succesfully a lot of parameters. An here it comes, would it be usefull? No, because the get all these parameters you would need quite a lot picture of the animal from all points of view , and that would be quite time consuming and cumbersome. So the human expert would win here from an efficiency of view from the machine, but on reliabilty, I'm not sure...
My humble thoughts...

Bernard, AI is now very advanced, it wins with humans in every game and is better than humans in many things. I think that it is possible that in future AI will identify species better than humans.

Bernard, what I said was:

Ilian Velikov wrote:I don't see how such a software could be 100% accurate, or at least more accurate than a human.

Meaning that I think the computer is less accurate than a human not that the human is 100% accurate.

Maybe there could be a way to give a computer so many parameters that it could become very accurate, although as you pointed this is not justifiable in terms of efficiency. However, contrary to you I'm not convinced about the reliability too. One of the main differences between how a human brain thinks and how a computer "thinks" is the way the stored information is retrieved. In the human brain all data is freely associated and retrieved with an astonishing speed (almost instantaneously). In the computer everything is in a mathematical order and the information is retrieved in a very different way. For this reason alone (not to mean there are no others) a computer would never be as reliable or accurate in identifying species, i.e. the human brain can notice subtle things such as silhouette, or shape of the head, or things like the environment in which the animal is photographed, the age, the gender, and it could make associations on previous experiences and gathered data to identify a species sometimes even if it's just a tail visible or a leg, or the gleaming shell of a turtle, or it's snout and nose poking out of the water. A computer doesn't have this "magic" that the brain has and it wouldn't have it any time soon because for it to have it humans have to give it to it, and humans can't give it to it because we still can't explain exactly and completely how our brains work.