29 comments

[ 3.1 ms ] story [ 72.6 ms ] thread
How does this compare to Seek?
I had not much success with Seek when it comes to plants. However, its variant iNaturalist is not bad for identifying insects if you take good photos and spend some time sifting through the results suggested by the app.

Flora Incognita is more like the PictureThis app with focus on the plants growing naturally in local ecosystems.

Seek actually just got a major overhaul a few days ago. Most other plant-identification apps heavily rely on citizen science data generated by iNaturalist anyways
This is a really excellent app, I use it all the time in my garden. I planted native plants, and as they sprout every year it's very hard to tell them apart from weeds at first. But Flora Incognito does a great job of identifying most plants correctly even as they're just sprouting for the season.
Really excited to try this out. I have installed Picture This in the past but they push heavily for a subscription which makes no sense for me when I only identify a few plants per year. Their business model is awful given that I suspect most people don’t have a regular need to identify plants.
AFAICT, it's neither open source software, nor open data.
I tried it out around the yard and it correctly identified a variety of native flowering plants (in and past their prime), a couple trees (one native to the region, one about 400 miles from its native range), and a few ferns. The only one I'm pretty certain it got wrong was misidentifying a ghost fern as a Japanese painted fern (I have both, they are rather easily distinguished). When I tried to suggest that the misidentified fern was a ghost fern, it was unknown to the app so I couldn't suggest it.
Did you try to suggest by the common name or the Latin name? It's highly unlikely that the app would have all the common names of all the plants. Even the common name "ghost fern" refers to a few different plants. Most commonly in the ornamental plant trade it refers to a hybrid variety called Athyrium x hybrida 'Ghost'

The other major issue with ornamental plants is that they're often selected for very particular physiological characteristics. This often leads to plants that visually look very different from their wild ancestor but are still extremely closely genetically related. If you've ever had a variety of a plant with variegated leaves for example, you'll probably trip apps like these up a lot. It's simply too difficult to keep up with all the new ornamentals being bred every year. The data just doesn't exist and they're quite often weird hybrids with uncertain phylogenetics anyways

Interesting...seems like a generalized issue with identification type AI?
Eh I wouldn't say that. You can selective breed certain plants to look extremely similar to other plants over the course of even a single year. As in, you would be able to trick a human being as well. If AI is making the same mistakes as a human, I don't really think it's a problem with the model as much as a limitation of the data

I mean taxonomists themselves have often been proven incredibly wrong once phylogenetic studies have been conducted. There's simply major limitations to identifying a species by it's physical characteristics alone. Not to mention you don't get to, for example, smell, touch, taste, or watch something over time. Really it's incredible how accurate they even are already

What about fungi? :(
iNaturalist has been pretty good for me with fungi and slime molds. You don't even have to save the report, just start by uploading your pictures and then looking at the suggested species. If you have multiple pictures it's worth moving between them, the classifier only works on one picture at a time.
(comment deleted)
I've had pretty good results with Google Lens in the past, but occasional misses as well. I'll have to check if this fills the gaps.
I've only used PlantNet in the past. Worked well for me the few times I've used it.

https://plantnet.org/en/

Whichever one you're using is probably borrowing massively from iNaturalist that also has (and has had for a long time now) its own AI-powered identifier (which recently underwent a major update).

iNaturalist is a platform to allow citizen scientists to upload observations of any species. The distribution data is what's most used in published research and databases like GBIF, but photos can also be uploaded with any license the observer wishes and iNatualist has decent APIs and works hard to make its data accessible to anyone

Recently discovered this functionality is built into iOS. Take a picture of a plant, view it, at the bottom see the "i" in a circle. Click on it, then a banner under the image will say something like "id plant". Works pretty well, not perfect but was surprised that this is a built in feature.
From what I understand this is more like reverse google image searches than a species identification AI. Regardless still pretty useful and often good enough
Come to think about it, image hashing isn't that far conceptually from vector embeddings.
Great app, but I hate their use of high precision GPS. Is it really that important to know the location of a plant down to meters??

If plant location is good to have, why not simply ask the user what area they are in (e.g. Berlin)??

I suspect it helps pinpoint invasive, rare, etc species and provides highly useful information for mapping distribution of species. Location is critical for understanding.

It should be entirely optional for passive viewing, but at least encouraged for active contributors.

That doesn't make sense.

You don't need the exact coordinates, just what region in the country you are in.

(comment deleted)