This is a plea and a promise
It is possible to test PlantNet in the test site, there are THOUSANDs of Observations.
Just use a Date Window
I would be willing to devote a LOT of attention to the way Plantnet ‘behaves’ None of us can do that in the live site. Our observing process is far too precious - honest
Thank you for this suggestion. @Chris_Valentine is not with us live today but will review and respond.
Thanks again to those who were able to join us LIVE in the chat and on Zoom. Further updates to to follow soon!
I found myself in the zoom but really wanted to make points about PlantNet
Silly me. Mike/Janice I am VERY keen to help the process IF you are going to integrate AI.
Phase 3 should be a test of the way it might be integrated
Phase 1 is complete - yes?
Phase 2, in a Test Zone, should be a continuation of Phase 1 and with BOTH parties contributing BUT the really serious errors of Phase 1 put right
Think this is an important point, we could try a range of possible integration methods on a test site as they would be very messy onthe main site, for example more/less suggested ID’s, scores and various ways of displaying these and ideally a button to switch off or on etc. However we would be constrained by the current software as to how much of this is possible to test. These suggestions have come up during the testing that we have been doing.
We are gathering all the feedback, hence why these types of discussions are such an important part of the process, and the next stage is to review this and further discussion by the iSpot Admin Team, to help us decide what next.
Reading back
This is an EXTREMELY garbled record because there are two main and overlapping themes
This is SHORT of regular, even irregular, Users (only three of us)
I am CERTAIN that if you began a normal test exchange here “Views Needed about PlantNet”
you would get a better response.
Yes, the delights of unthreaded forums.
The Pl@ntNet API has a daily limit of 500 images and a good percentage of that is already being used on the live site - to process the “back catalog” of thousands of plant observations would require separate scripts and would have to be limited to just a couple of hundred images per day.
Not sure which thread to post this in but noticed something on a couple of my obvs where PlantNet has suggested an ID. It’s regarding the common names. An example of this is that has suggested White Campion but lists a common name as bladder campion? I had a similar thing with Herb Robert where it lists a common name of G purpureum as Herb Robert.
That is interesting, Chris do you have a reason for this, is it just the list of common names includes many different common names that have been used over time and not just the currently used name?
Yes, that is interesting. The script takes the names supplied by Pl@ntNet but the link is to the iSpot dictionary entry and in this case, they’re not the same. Our dictionary has both Silene latifolia (NHMSYS0000463572) and Silene vulgaris (NHMSYS0000463584) as discrete entries.
Suspect something needs looking at here will email you separately about it Chris.
Thanks both - fyi this is the Herb Robert one
White Campion
White cockle, Flower of the Dead
Common names frequently overlap species
Many Common names are not in common use
Wiki sometimes gives a list of Common names
Woe betide anyone here who uses the Common name to generate the ID (Luisa didn’t of course)
Try Sycamore. Has someone mentioned this before?
.
The Thread monitor has just asked me "You’ve posted more than 22% of the replies here, is there anyone else you would like to hear from?"
What should I do?
Keep going!
.
I find the PlantNet is not to be trusted on common names, and common names a minefield. Never tried sycamore though.
Chris has now changed the code so Plantnet should now use the common name from the iSpot dictionary not their own (where there is an ispot common name for the shown species).
1 Like
That’s brilliant - thanks everyone.
Yes, once again where there is a will, there is…
Thanks
The 1410 PlantNet intervention was added yesterday, so it looks as though it might have been accepted as a permanent feature of iSpot