iSpot has a reputation system which ensures that identifications of say plants are more or less as accurate as from experts (based on a test of thousands of ispot observations that have been verified by UK leading experts).
2 Likes
Thanks good to know. Can they end up in biological records then, and if so how? I notice the NBN logo everywhere on the site and they must also therefore believe in the competency of the system?
ISpot is for anyone interested. Some of us are beginners in some fields and more knowlegeable in others. Some of us came with a biological background, others of us with very little knowledge.
.
All that is needed is an interest in learning and sharing what we know, or want to know, about the natural world about us.
2 Likes
Plantlife question. Does it only add ‘comments’ to ispot posts without an id?
No, it ignores any identifications - it just looks for the “Plants” tag. Its the same with the FASTCAT-Cloud AI - that just looks for “Mammals”. Both only operate in the UK & Ireland community.
At the moment, the two systems are called manually but eventually there will be a scheduled script that will call them daily.
Thanks. I would still like to know if the observations can end up as biological records on the NBN Atlas?
Some are already, we were hoping to get many of the others onto NBN this summer but have hit a small issue which I am currently working on.
1 Like
Thanks you Chris (plus characters to 20)
Question ( well you did encourage us to take part) . In this post
There is a discussion on the species ID. Thymus is agreed and plantnet * (PN) thinks so too. However I’ve suggested PN could stop at the genus. Does PN always offer a species? Is simply offering a genus not an option it is programmed to offer?
EDIT. * PLANTNET (PN) sorry for the confusion.
That is an interestig question JoC, I wonder if any of the plantnet people would like to respond on this?
Could we confirm here please that Plantlife (PL) here is a mistaken description of Plant Net and that we mean Plant Net please? I am getting confused
Could you please tell me how the records end up on NBN Atlas? Do they have to be sent to Local Record Centres to be put on the Atlas? I know that one can submit records in bulk and there is a you tube video from NBN on this but can ISpot colleagues do this? I am asking as I wish to understand the verification process…at some point it is determined that the identification is acceptable. Thanks Paul
The AI system is Plantnet Home - Pl@ntNet If you make lots of records and you are experienced at identification then you can submit records to irecord in bulk and an expert will eventually verify them but they will need to have some idea of how good you are at making records and the verification may take years.
If you submit records to ispot then they need to be one at a time preferably with at least one image for each observation then the community helps to check the observations. If it is something relatively common then people will agree very quickly.
Most of the ispot data does not currently end up on NBN, it was due to be there by this summer but there is a technical issue holding this up at present, hopefully will be resolved fairly soon.
Miked says “ at least one image”. If it’s common one image may be enough… but if it’s a plant or bee with several closely species, then one image is unlikely to show all the characters needed for us to agree.
The same presumably goes for AI ; it can only work with the information offered , which is I think, at present only images. In addition we use other clues like location and habitat. AI could do that, of course, when it’s programmed to do so.
Yes, absolutely correct and this is one reason why we need to raise the awareness of our use of the API - to get users to understand how important the images are. The mammals system (FASTCAT-Cloud) currently only works on a single image, whereas Pl@ntNet can take as many as five for a given observation.
I think Plantnet have a system that can take into account multiple factors based on the location and what it knows about that place from standard environmental datasets (not sure if it also allows you to give it specific other typed in info) but that is not in the main Plantnet AI system.
In the interests of trying things out, I was thinking of making a post of a plant, Veratrum, that some people mistake for Gentiana lutea. If they do they get poisoned.
.
We could see if plantnet spots the difference.
.
It’s not British, so I’d have to post it in a False Location for the test, say in Buckingham Palace.
.
Would that be in keeping with the spirit of this AI project?
iSpot only submits observations to Pl@ntNet from the UK & Ireland community but otherwise it won’t care where you put your observation. And the API doesn’t take any notice of any existing identifications either - it only works on the images.
If you want to give their system a better chance of identification, you can name your images according to what part of the plant they show - the tags it looks for are flower, fruit, bark & leaf. For example, if make the word “flower” part of the filename of your photo of its flower, when my script submits the images it will add that tag. And as Pl@ntNet takes up to five photos, you could include one each of flower, fruit, bark and/or leaf as well as an overall photo.