when I was in school we had something called "dychotomic keys" which were big books to recognize plants with binary questions.

Does it have a leaves? yes/no go to page XXX/YYY Is the leaf oblong? go to page WWW/ZZZ Is the border smooth? go to..

I wish there was a plant identification app using _that_ system, it would be much more entertaining and educational to use.

For woody plants, VTree might fit what you're looking for. It's a dychotomic key optionally augmented by your location data.

https://dendro.cnre.vt.edu/dendrology/vtree.htm

Wonderful, thank you for sharing!

I've been meaning to take a crack at a software version of this for insects/birds/plants, do you know of any good ideally libre sources for the data?

I'd be willing to help you. The largest dychotomic key [1] is partially digitised but we could transform it into a proper dataset. Copyright no longer applies to part on the Flora. There are many other sources. I remember dychotomic key in hypercard [2].

Scanning 81 million science paper pdfs will yield most of the data in [1] and [2]. It would be possible to get a grant to make these transformations and add them to Plantnet and iNaturalist.

[1] https://en.wikipedia.org/wiki/Flora_Europaea

[2] http://www.etibioinformatics.nl/faq/

I am away from home atm but very interested in making this happen. I'll look into it a bit and get in touch by week's end! Thanks for the links.

It is also possible to add DNA material to field observations. A DNA scanner cost around $1000 and attaches to laptops and cellphones so they could be used in the field. I'm not sure if amateurs can handle the software yet, but that could be fixed.

The most progress could come from my speciality: data mining scientific papers. This requires a few hundred terabyte of harddisks and a fast computer plus 2000 hours of programming. This would require a minimum of $20K or more in grants, crowdfunding or donations but it would yield an enormous boost of data.

Another addition could be scanning all the herbaria and botanical gardens, both photo's and DNA samples. Crowdsourcing by thousands of citizen scientist would be the way to go.

I've downloaded almost 2 million photo's from several bird, insect and plant databases this weekend and (partially) compared them with the Plantnet and iNaturalist datasets. And I looked at the dozens of other databases with mosses, lichen, bacteria, etc. Still, no where near the >8,7 million species of the world are in those databases. The most complete list would be the plants of Europe (only a few thousand out of 400K species), most other regions and kingdoms are only partially identified. As we can see from dozens of papers, new species have been found through the iNaturalist collecting. Making better software and more data would certainly boost the rate of new species discovery. And that is vital, as we are in the middle of the 6th mass exinction event of the last 570 million years. Species go extinct before we even have photographed or identified!

I also have my dad's Flora research photo's from 30 year field trips and my own, of which 3000 plants where identified with Flora Europeae dychotomic key (a day's work per flower).

Yes, I know of quite e list of libre sources, only some of which are digitised and online. So (continuously) exporting data from the hundreds of small observation or collection databases or going our and scanning the old offline materials would be what you could organize with better software.

There are also ways to increase the collection of libre data. You could auto-generate field trips for holiday makers or school trips. Give them an auto-generated itinerary leading them past places with flowers, insects or birds with a list of things they should look out for and photograph or determine with keys. Make it into a game like geocaching or a treasure hunt. You generate it based on individual tastes (walking, bicycle, car tour, camper, bustour), individual or group size, age and knowledge level, temperature, season, time of day, climate, location, etc.

The US Department of Fish and Wildlife has a really useful service called the Feather Atlas for identifying bird feathers by their color, size, and shape:

https://www.fws.gov/lab/featheratlas/idtool.php

That's correct, there's plenty of incorrectly identified plants, but the app says 89% accuracy for the first result and 11% for the second, so it's hard to argue. I've many times encountered plants that don't in live in the UK but the app say they do.

Inaturalist uses GPS and date to help with this. So it shouldn't show you stinging nettles in November.

The high accuracy results are not always, well, accurate. Several times I've had an 85+ percent result then looked closer and gone "ah shit, the leaf pattern is wrong, these are alternating instead of side-by-side" or similar.