Bulk Selection of Observations to Add to a Project

Thanks, Dave :-)

I personally would still prefer bulk edit and add features rather than letting AI loose on it. Perhaps I’m just old fashioned? Or maybe just old ;-D

1 Like

I dunno… it’s much less “work” to just be correcting a few AI mistakes than having to do the whole lot manually. Even if they are checked in their entirety it would still be approx half the workload… but I do question whether we NEED it done, or just IT WOULD BE NICE TO HAVE IT DONE. Maybe if they come up with some onboarding tutorial thing like they do with the forums, they could have the annotations part of that taking the new users through how to apply annotations to existing observations, with the lepidoptera being an obvious easy training case. Each new user gets to do 5 lep annotations each before moving on to the next task… could also ingrain a practise of applying the annotations on new observations, reducing the future workload of this!

If you made the project a collection project, then all you would have to do is annotate observations. They would then automatically be included in the collection project. Or am I missing something?

2 Likes

Tony - I see two issues with this. There is the ongoing challenge of accessing details of obscured observations in collection projects, so that doesn’t work for many people (speaking as a data user).

The second issue also addresses karoopixie’s comment. The math is against manual annotations and I don’t share your belief that waiting for people to manually annotate millions of photos is realistic. I’ve posted elsewhere that of the over 5+ million insect observations on iNat, only 20% have annotations. There is no evidence that it will get better. That obviously excludes the other possible annotation categories, but I can’t imagine things improving much.

2 Likes

Tony, yes your suggestion would work if all obs could be annotated. Unfortunately, the rearing obs which have all life stages cannot be annotated, and these are the most valuable ones for identification because it has the larva and adult confirmed.

And, yes, I am well aware that iNat only wants one individual in one life stage per ob. But, hundreds of rearing obs were migrated from iSpot, and the posters are not going to split them all up again. Some still prefer to upload rearings as a single observation, and I can’t say I blame them. It is a nightmare trying to link all the separate obs for one rearing together.

I am thinking… Let me do a little experiment, and I’ll get back to you with the results…

1 Like

Okay, I have done my little experiment.

I created a collection project for the same params as the trad project. These are the current totals:

Numbers          Tradpro      Collpro
Observations     4505         3794
Species                387          312

This is mostly because people automatically add to the project without adding any annotations, and because the rearings are not annotated.

The projects are here if anyone feels inclined to see them:
Trad project - https://www.inaturalist.org/projects/pillar-parade-s-afr
Coll project - https://www.inaturalist.org/projects/pillar-parade-collection-test
I will be deleting the collection project soon as it was just done as a test. You will also see anomalies in numbers of obs outside the southern Africa place because people have randomly added those on the trad project, but all are added on the collection project (and I wouldn’t necessarily include all those places on a coll project since it was supposed to be just for southern Africa anyway but I included them for testing).

1 Like

I hear you. We were talking about your proposal yesterday and it is possible, although it would take some time to train. But as I said in that thread, just as our species ID AI was trained on observations with millions of IDs made manually, so would this AI, and the community would still have to go through and correct AI mistakes for the next training run to improve things. As to whether we have “enough” annotated observations now to train an acceptable model, that’s something we can look into. But more will pretty much always be better.

@karoopixie interesting experiment, thanks for sharing.

1 Like

@karoopixie

You have a trad project, and now a collection project. Create an umbrella project that picks up both, and then (after waiting enough time for indexes to catch up) do a search on leps NOT in the umbrella project. That will be your shortlist of what NEEDS to have annotations added. As soon as you add the annotations, they will appear in the collection project, and so will drop out of the umbrella project…

I’ve not tested it for annotations, but we do a similar thing to catch obscured records into our bioblitzes…

1 Like

@kiwifergus

Thanks for your suggestion. I had considered an umbrella project (though in a different way) but it does not solve the initial problem, and creates many projects.

I already have a way of finding the unannotated leps which is much simpler than your solution (which is not to say your way doesn’t work, and it may have applications elsewhere), which is to do a filter on the Identify page like so:
All sA leps: https://www.inaturalist.org/observations/identify?reviewed=any&quality_grade=needs_id%2Cresearch&order=asc&taxon_id=47157&without_term_id=1
Just sA moths: https://www.inaturalist.org/observations/identify?reviewed=any&quality_grade=needs_id%2Cresearch&order=asc&taxon_id=47157&without_term_id=1&without_taxon_id=47224%2C47653

I work on the second filter because as you can see I only have 8,000+ (of which about 400 are unannotatable) moths to annotate. I’ll leave the butterflies to someone else because I simply don’t have time or inclination to annotate 18,000+ obs (I have already annotated ~37,000 obs).

Which leads us neatly back to the original feature request of bulk adding annotations, and bulk adding to projects. This would greatly streamline the whole process, and eliminate the need to create multiple projects to get to the same destination.

1 Like

Cool… add the “not in project” filter to yours, and it might cut down the list a bit.

Ah, no, you see that is Step 2 of the process!

See my comment above:

The link for Step 2 is:
Larva obs not in project

that is not what I mean… you have a filter that gives you all the un-annotated leps, but some of those might already have been added to the target project manually. As you say, there are a lot of observers and identifiers that add larva obs to your project directly. The addition of the “not in project” filter is just so you don’t have to work through those when you are putting the annotations. Unless of course thoroughness is important, and they must ALL have annotations!

I can see why creating multiple projects is a pain, especially when you have a base of followers that are usually accessing the project through one project page and they would likely have to start looking at another (the umbrella project), but that doesn’t make your way much simpler than mine. Your step of then having to go through and manually add them to the project would be completely eliminated by the collection project being in play. If you had 50% larva obs, then that would be 25% off your total workload. How is that not easier? It sounds very much like you are quite attached to your process!

But yes, bulk add and bulk annotate would definitely be easier, and wouldn’t require the extra projects.

I’m in NZ so it won’t affect yours, but I do try and add life stage (and sex where determinable) for lep obs in NZ.

There was a topic started recently, where you could nominate a taxa, and give tips etc on how to ID, and then volunteers could join you in working on those IDs. This would be along those lines, you could post the filters and have volunteers doing the annotation side of it. Can’t think how to find that topic at the moment though…

[edit] Found it!
https://forum.inaturalist.org/t/help-me-identify-non-experts-welcome/2915

1 Like

Thanks for pointing out that I may be attached to my process - we all need a kick in the butt now and then ;-D

I created an umbrella project just to see what we’ve got so far:
https://www.inaturalist.org/projects/pillar-parade-umbrella-test?tab=species
I don’t find this particularly helpful (but then it is afternoon and I haven’t had lunch yet, so it may look more rewarding in the morning).

Yes, I do want all annotations added because I think the graphs on the taxon pages are very useful. I just quickly got these 2, which are maybe not really representative of what I want to show, but they’re interesting nonetheless:

Hippotion celerio
hipcel
Hippotion eson
hipeson

This is fantastic data which can tell us flight periods, relative longevity of life stages, number of generations in a year, differences between similar species etc. So, yes, maybe I’m a bit OCD about adding ALL life stages ;-)

Anyway, I think if the following three features are added, especially the first one, it will go a long way to making the whole process faster which will encourage more people to add annotations/projects and take the load off the 3 or 4 who are doing it now (for this particular project):

Thank you for adding annotations to the NZ leps - I know what a big job it is, but the rewards are very great.

I will have a think about asking for annotaters on the other thread you linked (after lunch).

I do appreciate your help and suggestions very much, even if I don’t necessarily follow them. They always do send me down an interesting road though, which is fun :grinning:

1 Like

I totally get the CDO thing… :)