Back in the days of Sinclair, there was a book on BASIC programming titled ZX SPECTRUM BASIC Programming. The Internet Archive has it, in both web-browsable and PDF format, which is excellent.

One of the example programs in the book was called Pangolins - where you think of an animal and the computer asks yes/no questions to determine what it is. After a few questions it makes a guess and (if wrong) asks you to provide a question to differentiate your choice from its guess. Being less than a hundred lines of code and a very limited dataset it was obviously crude, but I guess it must have been one of the first bits of software I typed out for me to remember it so many years later, (or perhaps the memory was helped by the juvenile fun of entering sibling names alongside insulting questions).

Anyway, so we've now got this Internet thing; countless different databases describing in detail all manner of living creatures; huge amounts of hype pretending "machine learning" is a wonderful new thing - and yet there appears to be no decent website for narrowing down unidentified life via the simple process of answering distinguishing questions.

Why the fuck not?! :@

And so the next tech unicorn was conceived: Critter Twitter.

Gofundme, or SoftBank?

What's stopping the likes of Amazon/Google/Wolfram from throwing resources around and making any such efforts redundant?

I dunno. A patent troll?

:|

What critter are you trying to identify? Got a picture of it?

8 true flies, 7 dragonflies, 1 damselfly, 3 grasshoppers, 4 beetles, 2 pondskaters, and a pheasant in some grassland.

Also at least a dozen different fungi, where being able to filter by shape and colour would be a nice starting point.

doesn't google do an image search thing (you upload an image, and it searches for similar) ?

Also, there's this https://merlin.allaboutbirds.org/

and this

https://www.insectidentification.org/bugfinder-start.asp

Technology has moved on, my friend.  (wrong)

Google's Image Search has improved a huge amount since I last tried it - now it actually returns living creatures of similar shape instead of just photos that are the same colours - but it still gives a lot of wrong answers and doesn't say anything about the species.

Merlin appears to be mobile-only spyware.

BugFinder is an overly crude tool that only works for North America. (Though it did tell me one of my beetles looks like a Soldier Beetle.)

https://uk.whatbird.com is the least worst bird identifier I've found - it's often frustratingly slow and not easy to use, and its example images can differ from what an image search returns, (so potential matches need to be checked twice).

The general problem with these magic AI image matching services is their assumption of infallibility - they don't tell you which criteria they use and allow a human to tell them which ones are correct and which should be discounted.

For example, Wolfram's answer to one image (containing a hoverfly) is that it's an "Africanized bee" - despite that only existing in South America and southern US, (not to mention the image quite clearly being a fly and not a bee).

Google's answer for the same image is a Short-tailed Blue, which at least gets the correct family of the butterfly, if not the right genus (it's actually a Holly Blue).

(Side-note: If I didn't know they were both wrong, the fact that the Short-tailed Blue isn't found on the American content would be evidence that one of them was incorrect.)

Neither Wolfram nor Google gave any hint of recognising more than one thing in the image. Neither of them get me closer to identifying the pink-flowered bush the two insects are on.

Have you tried crowd-sourcing the ids on insect|bird forums etc?

On Monday I photographed over a dozen different species; I need a more scalable solution than uploading images, asking people and waiting for a reply.

My ideal solution would be an open source Lightroom plugin that did image recognition for the basics, then asked questions for the bits it's not certain about, and worked fully offline.

What I expect I'm going to have to do instead is spend a significant amount of money on a stack of books that will then become a constant source of getting in my way.

Books are underrated, these days. Go with the books.

You tried the library?

Wolfram thinks this is a Lycopodium obscurum.
Google says Juniper tree.

If I could tell them it was found in Pembrokeshire woodland they'd know those answers were wrong.

But I can't, so I'm back to guessing unknown attributes and scrolling through an endless haze of mature pink flowers trying to figure out if any of them match what this bastard thing might become.

:@

"This" being a photo that refuses to attach, which probably means there's something wrong with thumbnail generation.

Holcus lanatus

Commonly known as Yorkshire fog, (plus a bunch of other names). It certainly looks like that, and is supported by Wikipedia's assertion of "Holcus lanatus in its natural habitat is a food source for butterflies such as the speckled wood..." - because that's what the flying caterpillar sitting on it is.

Thanks. :)

I used the "Picture This" app for that.
It only gives a few free searches though.

And appears to work with a browser except the upload icon just prompts to install the mobile app. :@