Pl@ntNet Confusion?

1

I used Northern Europe: Identify - Pl@ntNet identify to help me identify a plant (see https://www.ispotnature.org/communities/uk-and-ireland/view/observation/864747/horse-radish-armoracia-rusticana). The first 4 suggestions offered were
Horse-Radish 9.07%
Greek dock 2.53%
Caucasian beet 0.84%
Russian dock 0.74%

I see that Pl@ntNet-API has commented on my post and offered
Greek Dock (score: 0.34)
Horse-radish (score: 0.13)
Russian Dock (score: 0.06)

Can anyone explain the differences in

  • the scoring systems used (I guess that a “score” of 0.34 from the API is equivalent to 34%??)
  • the scores between the two versions of Pl@ntNet
  • the capitalisation of the common names (reproduced accurately above; it’s not just my mis-typing!)
2

This has nothing to do with the issue you raise, but…
I went to have a look: the Gremlins are having fun again.
The post, as shown in the list view, doesn’t display the image. When I tried a screen-shot to illustrate, it wouldn’t paste:
image
If I select the observation, the photos do (eventually) load. But when I go back to list view, nah!
I took another screen-shot and posted it into my image editor: here it is -

Untitled-1
Untitled-11111×996 102 KB

3

I think the plantnet site itself does work differently to the API as they are constantly tweeking and it is easy to get out of synch i.e. one bit of code more up to date than another so I am not surprised to see a difference.

4

But some quite significant differences. Any why do they seem to have different dictionaries (at least for the common names)?

5

A possible hypothesis is that iSpot gets the scientific name from PlantNet, and looks up the common name in its own dictionary.

6

Yes I think that is the case, it may even be just the number and getting the scientific name as well as common name but better Chris replies on that part.

7

Absolutely correct. Pl@ntNet sends a field called ‘scientificNameWithoutAuthor’ which I search for in our scientific names table to get the NBN key. From that I retrieve the recommended common name from our vernacular names table.

8

Have just fixed the issue with the Pl@ntNet AI calls not working. You should again see suggestions from their system in the comments on plant observations.