Snake ID app! I need your help!

[adela]

T Hi snake lovers! I need your help!
I’m trying to make an app to identify snakes using photos. I’m interested in your opinion on what the application should contain. Please fill out the questionnaire and I will let you know when the application is ready. Thank you, it helps me a lot!

Questionnaire here.

[eaglereptiles]

Best of luck to you with this project, though I think you could use some help wording your questions and options for your questionnaire.

@chesterhf took a stab at this a while back…

Machine learning and Morph Identification - It actually kinda works

So I’ve been working on building a machine learning program to identify morphs just for fun, and to my surprise it’s actually not half bad. It’s been trained on ~350 images so far using only a handful of basic morphs (Normal, Spider, Mojave, Leopard, Butter/lesser and Clown) and is currently running at a combined accuracy of 96%. Current guessing accuracy: Clown - 98% Leopard - 95% Normal - 100% Mojave - 91% Lesser/Butter - 91% Spider - 98% Made into an app, this would actually be pretty…

With some pretty decent results

@chesterhf

Current guessing accuracy:

Clown - 98%
Leopard - 95%
Normal - 100%
Mojave - 91%
Lesser/Butter - 91%
Spider - 98%

But I suspect the more morphs (and eventually combos) you add, the lower these percentages will get.

[adela]

Thank you! That is interesting.

[auriea]

You should be able to put in the dam and sire morphs ofc. The only issue would be with unknown hets and morph redundancy. Like banana and coral glow, or morph stacked names, like the Bee complex being spider+pastel (+whatever else) or the BEL complex… Being the bel complex, lol.
It’s not impossible to teach ai all of that jargon, but it is subject to frequent change.
Allowing the ai to account for hidden hets could be beneficial, but it also could allow the ai to go crazy with whatever it sees at random in the animal.
And of course it would have to have a solid understanding of incomplete dominant and homozygous and all of that.

[adela]

Thanks for comment!

[verinium]

Take this as friendly advice, and not criticism, but what it needs above all else is a disclaimer saying it is not designed or intended to be used as an expert opinion regarding identifying snakes, and is only intended as a “best guess” software. There are many snakes that look EXTREMELY similar, and i doubt an app would get it right every time, this could lead to somebody getting bitten by a venomous snake due to false confidence brought by treating the apps response as 100% definite.

This should prevent lawsuits and also remind people that an algorithm is no replacement for proper knowledge and familiarity, but it is better than nothing if used to sate curiosity.

Just an example, can you spot which is the venomous snake of these 2 pictures? And are you confident the algorithm could do the same?

I think its an awesome idea as i myself use identifying apps for alot, its just important people know the apps limitations xD

[verinium]

Reading the other comments it has come to my attention that by “identify snakes” that you meant their genetics and not what species of snake it is xD.