there are now 235 interactions from P@N.
I think it would honestly be nice to browse them. I have invented tag - did I say?
plantnet one word, lower case. Click on it to see who is responding. I do hope no-one redevises the tag.
Use Community Explore to find them if you don’t have one to hand (there are only 7)
Many of us will find that Plant@net has been to our Observations - there is NO way to tell without looking. I am finding it VERY frustrating not to be flagged in my Changes tracker
PLEASE add the tag, please ask for the tag to be added
.
Spot-on here but it shows how we might have to modify our presentations, making the lead-photo one of the Observation subject
I have established that I would need to add records to a couple more tables in order to flag Pl@tNet or FASTCAT-Cloud comments in peoples’ Changes Tracker. Our concern is that for those who very kindly post a large number of the relevant observations it might become overwhelming. What are peoples’ thoughts?
As I understand from a previous comment (here or maybe elsewhere) plantnet comments are being added to new posts of mammals & plants. So if I post one of those (well, realistically a plant) then I would expect a plantnet «comment» within a day or two. So for me, an alert is not high on my Wanted List.
.
I don’t think I am considering adjusting my current mode of posting… with the order of photos reflecting the features that I think most useful for the ID I am proposing.
.
I like AI, but I’m not adjusting my practice to please it UNLESS I have no idea what the plant is and SEEK has no idea either.
.
But of course, I may change my mind about one or other or both of these observations later.
hornbeam two for price of one | Observation | UK and Ireland | iSpot Nature another interesting example of AI getting things wrong and reasons why it has failed and why humans are better in this case
Plant net is not wrong here https://www.ispotnature.org/communities/uk-and-ireland/view/observation/849240/
But my photo fooled it
.
P@N probably needs to have expanded code to subdue or preclude species not occurring in the UK but ONLY if it is the UK Version (of P@N)
.
The references should NOT be to the iSpot Dictionary but the UKSI one which is being updated continually.
@dejayM @miked
Okay so I’m fed up now, losing the will to add plantnet tag to plant observations! The weighting seemingly given to non-UK/W Europe species makes it less useful IMO than the app where flora of W Europe (and elsewhere of course) can be selected. And it takes up such a lot of page room on a smaller screen.
And I see @Chris_Valentine ‘s point about overwhelming the tracker, since it looks like all plants will have that input as a matter of course during the trial?
Don’t worry - I’m modifying my scripts to add the tags as it adds the comments. I will also retrospectively add tags to any that already have the comments. I’m doing this for both Pl@ntNet and FASTCAT-cloud.
The code should add the plantnet tag from now on so you should not have to.
Will discuss with the plantnet people about which species are being searched for, they have changed their system many times so there may be options for choosing specific regions. However quite often the species that fail to get identified in UK are garden plants or escapes so could come from anywhere and in this instance having a much wider range of species to choose from might be useful
Thanks Mike. Yes of course garden escapes etc need that wider search. A friend and I had some fun in a park yesterday with Seek (her) and Plantnet (me). See my comment, if you scroll down far enough, at
That big block of stuff is now annoying me !!!
But why use world flora as the default initially with UK observations also (see “Reminder” grass observation linked above) why give poss IDs which the app declares to be only small percentage likely?
These are good questions which hopefully the developers of plantnet will be able to answer. Hopefully they will look through these comments and respond but have been ill and on holiday so not had input from them directly so far.
Thanks, hope all well @miked
iSpot has over recent years become very rich in learnings, details and comments as well as in IDs - although for many hopefully happy-contributors/observations, a photo with an ID is all good of course! For me, it’s important that contributors are involved in considering such a major change, should this be proposed as a permanent feature for the site (as the title of this discussion may imply).
At the moment, my concerns are principally:
•AI suggestions for UK obs should have geographical focus in the first instance rather than use the world flora by default as it appears to do now
•Giving suggestions to which the app (when same image used) allocates a low probability is potentially confusing
•Panel is huge, imposing and distracting esp on small screens - clutter! I am not liking it.
As I say, I am struggling to see the advantages of this proposal.
I haven’t yet used PlantNet though I thought I might try it (don’t use a smartphone and haven’t gone through the pages on the pc yet). That statement in itself might indicate I don’t know anything about it and that’s actually the case.
However, I must say I am now highly irritated by the blocks from PlantNet that crop up in many comments in the observations I look at and even more so when the ID confidence level is “as sure as I can be”.
One of the principle aims of iSpot was to help contributors improve their identifications skills, how is using AI assisting that goal. The suggestions come in late and are sometimes obscure and are adding confusion to the site. Go off and research the subject yourself, compare it to other posts and then come to your own conclusion. That is what the site should be about. We still don’t have a next and back button on the carousel following the awful recode of 2017, maybe that should be addressed first.
Yes, the lack of geographical focus seems to be a problem with other AI systems too. I’ve been playing with Google Lens recently, mainly with moths. Too often it comes up with an American suggested species for a UK moth. When it does suggest a UK species it is often usefully accurate, not necesarily to species but sufficient to give a useful starting point for a book-bases search.
emphasized text**emphasized textI think the feedback we are providing is exactly what ispot is hoping for.
.
This from Janice on the issue:
“Cos4Cloud is a European Horizon 2020 project supporting citizen science by developing technological services to address challenges shared by citizen observatories and help them boost the quantity and the quality of observations. The Open University is a project partner and iSpotnature.org is a Cos4Cloud citizen observatory and this role includes integrating and testing relevant services. iSpot is currently trialling two Cos4Cloud image recognition technologies FASTCAT-Cloud and the Pl@ntNet-API . We need YOUR input: This is a Cos4Cloud iSpot User Group set up facilitate discussions, gather feedback, etc. from the iSpot Community as you use and test these Cos4Cloud services.”
I also believe that P@N’s presence in its current form is very distracting,
I have been going out of my way to be supportive but am not happy that the site might become encumbered with a dimension that is not required.
I entirely agree with @chrisbrooks about the need to improve the way iSpot currently runs, before adding NEW facilities, like, for example, Tweets and Live Chats.
Please address the ongoing concern about currently failing aspects of smooth running and the apparent almost daily loss of users.
Yet, possibly thanks to P@N, I have just offered a ‘first for iSpot’ Observation here https://www.ispotnature.org/communities/uk-and-ireland/view/observation/849553/
Still only 32 with a tag (out of 252) here https://bit.ly/3PAJAe7
Chris is still working on the code, they should all be tagged eventually.
Yes, it’s reassuring though that some users are reading my small print.
Some of P@Ns contributions are wild to the point of off-putting.
https://www.ispotnature.org/communities/uk-and-ireland/view/observation/849449/
My experience in iNat is that new users are quite likely to accept first suggestions.
Here in iSpot we might not add comments or additional IDs, who wants to spend time on errors made by AI?
My case is supported in The Biology of Survival Observations which were often to Species and probably from AI, quite few of them have no supporting agreements or comments - ignored!
https://www.ispotnature.org/communities/uk-and-ireland/view/observation/801243/
https://www.ispotnature.org/communities/uk-and-ireland/view/observation/799894/
https://www.ispotnature.org/communities/uk-and-ireland/view/observation/761321/
etc
