We would like to thank you again for participating in this chat. As we said last time we need your input giving us feedback, raising questions while testing the PlantNet-API as well as FASTCAT-Cloud in iSpot.
As was noted, the AI is working in one direction in that it pulls an observation and suggests options in the comments section. We need human computer interaction i.e.you interacting with both services - such as adding comments and responding to the AI suggestions: were the options useful or correct? This will help to give us feedback on how they are performing.
This is the second iForum LIVE! sessions to facilitate scheduled group discussions Two others will follow and we are proposing using Zoom as part of this, , as well as the LIVE forum chat:. Please fill in the poll above to let us know if you would be interested in this.
Thanks again @ajoly and all members of the iSpot community for participating.
Iām no good with bryophytes (I record 3 species of liverwort and one of moss, while I manage approaching 1000 species of vascular plants in the wild per year), but the impression I get is that the distinctive groups are at ordinal level.
It was quite a different experience for the few of us who were in the parallel video conference to this forum.
It was easier to get more detailed answers and discussion with Alexis i.e. he often mentioned more things while he was typing his answers into the forum - speaking is quicker than typing. One of the questions I asked verbally and he replied verbally was about bounding boxes i.e. does Plantnet put a box round the object to be identified in the picture. The answer is no, this is quite a big difference to the other AI systems we are trialling on iSpot which always put boxes round the item(s) to be identified and which can give identifications for several objects in one image. Plantnet are looking into whether putting a box round the item to be identified would improve things.
Zoom and various other video conference software is free for the user, having a camera is optional, the main thing is being able to listen and might want to occasionally speak. We might set it up so that there is a parallel iForum live so that people can still type there and will keep an eye on questions in that place too. In the session today it was just the iSpot team and Alexis in the video conference but it would be good to have this open to all ispotters too.
Hereās one that Iāve just come across - while itās a sufficiently odd (compact) honeysuckle that I spent some time wondering what it species it is (keys out as Lonicera periclymenum in Sell & Murrell, but I was wondering about some horticultural form not in the Flora) itās obviously a honeysuckle (and asterid) and not a caryophyllid (Eriogonum) or rosid (Hyp
And an example of it failing on Urtica dioica. (One can understand picking a different nettle, and even Hesperocnide, but Sphaeralcea has very different looking flowers.)
And a misidentifed alder (with a fern as the second suggestion, which baffles me)
I can understand why it made a mistake with the āhunting tripā in this case having the search boxes shown (when that feature is developed on Plantnet) would help to see what it is doing. Just trying to match this overall image may get nowhere due to all the people and very mixed up tree but if it could search for leaves and put a box and try to identify within that then it may do better.
With the nettle, the image is not very sharp and could be one of the other species that Plantnet suggests, in this case having the geographical area (and saying it is not a gardent plant) may improve things a lot.
The honeysuckle it does get but not as the top suggestion. I canāt remember if it looks at more than one of the images via the api but looking at several of the images may help here. as you mention it is not the most typical growth form for the honeysuckle but I have certainly seen it growing like this especially on dunes and similar low nutrient habitats.
Oh dearā¦ I came and left a message (which has been removed!) just before opening time, but the event had ben cancelled, so got on with the rest of my day
I am not finding it easy to trawl back through the above
I WAS planning to ASK why no-one had opened a Project
.
I have done that now but without consultation
I too came to log in Wednesday with 10 minutes to go.
.
There was a message saying it was cancelled. Though that message seems to have disappeard.
If the message had said it was a technical problem and suggested we hang around I would have done so.
.
Iām disappointed.
So now I have trawled through this forum.
.
Firstly, thanks for answering some of my earlier questions.
.
Stepping back in timeā¦ did iSpot ask Plantnet to get involved? And if so why?
.
Or did plantnet ask to have access to ispot for their own purposes?
.
The answes may help us to see what the whole plan is and how we might best respond.
.
Stepping back in timeā¦ did iSpot ask Plantnet to get involved? And if so why?
iSpot (strictly speaking the OU) and Pl@ntNet are partners in the European COS4Cloud project and the use of AI in citizen science is a part of that project.
Is it reasonable to run the next phase in the Test Site, where genuinely interested people could help the development.
I have been through most of the 1044 Contributions and noted that few have comments AFTER P@Ns contribution. This suggests that is little feedback from mainstream users. Only three (Impartial?) mainstream users attended the Forum yesterday and just a few more the first one.
I have left comments for less than a quarter of users who probably have no idea this is going on.
I had answered the issue about seaweeds etc in your own project /www.ispotnature.org/communities/uk-and-ireland/view/project/852902/plantnet39s-contributions
You might have noticed this one - Nothofagus, Quercus and Fraxinus offered for a probable Populus, and only the Nothofagus has a ready explanation (similar bark in photographs without good views of leaves).
I too saw the message that implied the live forum was cancelled so did other things. I use app frequently, usually as a tool as to where to start looking in books and the net etc. Often I used to go page by page through a book, a soothing activity with a glass of wine. So I am not against AI of Intelligence of any sort in fact. However, I think the brutal way that oversized comments have been added does no favours to the whole concept. Particularly as the comments seem to offer no obvious feedback route, so I could not see what the purpose was. Not including the percentage likelihood is really annoying as that is a key feature when I use the app. A much smaller comment with three suggestions and their percentages with a link to rate the suggestions might gain some useful feedback . There should also be some sort of restraint in adding a comment if the three suggestions have a very low percentage and are from completely different genuses. Or a comment saying the AI has no idea, and a maybe a link asking the user for additional information.
Sorry about ramblings, but currently away so donāt have laptop to hand for proper editing with a decent screen
to let us know if you would be interested in this.