ID Tips for new moth'ers (10 Viewers)

[Surreybirder]

There are thought to be about 160,000 species of moths many of which will not have been photographed. No doubt the AI will improve but there could be lots of rather similar looking Noctuidae for example, some of which the AI may not have any visual data on. I am not sure that the AI on iNaturalist looks at the location. If it does, it comes up with some very odd provisional IDs.

 

[pppppvvvv]

There are thought to be about 160,000 species of moths many of which will not have been photographed. No doubt the AI will improve but there could be lots of rather similar looking Noctuidae for example, some of which the AI may not have any visual data on. I am not sure that the AI on iNaturalist looks at the location. If it does, it comes up with some very odd provisional IDs.

Well yeah, if there is a species that hasn't been photographed, the AI is not going to be of any use.
You'd have to fall back on the (often very knowledgeable) people who verify identifications on inaturalist to assist.

 

[pppppvvvv]

Inaturalist grabs the location off the picture metadata, or you can add it manually, which is a bit of a pain .
(My camera gets the GPS off its phone app, so I need to turn the phone app on, before I go shooting pics. Guess what? I always forget!).

I'm not sure if it factors in habitat. It should, I mean it wouldn't take that much extra code. A sighting on the coast has more in common with the coast 20 miles away, than the broad-leafed woodland 200metres inland.

 

[Richard D]

Inaturalist grabs the location off the picture metadata, or you can add it manually, which is a bit of a pain .
(My camera gets the GPS off its phone app, so I need to turn the phone app on, before I go shooting pics. Guess what? I always forget!).

I'm not sure if it factors in habitat. It should, I mean it wouldn't take that much extra code. A sighting on the coast has more in common with the coast 20 miles away, than the broad-leafed woodland 200metres inland.

Yes - having location set prior to doing the 'what have I seen' bit does improve te accuracy significantly. The AI isn't perfect but with two or three photos from different angles I find it pretty good - generally gets it right from one photo, or comes up with reasonable suggestions. I don't know about habitat - how well is that mapped?

 

[pppppvvvv]

Yes - having location set prior to doing the 'what have I seen' bit does improve te accuracy significantly. The AI isn't perfect but with two or three photos from different angles I find it pretty good - generally gets it right from one photo, or comes up with reasonable suggestions. I don't know about habitat - how well is that mapped?

The data must be available to some extent, UK wide at any rate. Chalk Vs Acid, broad-leafed or coniferous woodland, wet Vs dry!

Edit: even by reverse engineering from typical species solely associated with a certain habitat. I think AI should start showing us exactly how intelligent it is!

But I suspect inaturalist only relies on other sightings from the same area.
So if you are running a moth trap in the only small patch of coniferous woodland, on private land never visited before, surrounded for miles by broad-leafed woodland, you might get less accurate suggestions .

 

[Richard D]

The data must be available to some extent, UK wide at any rate. Chalk Vs Acid, broad-leafed or coniferous woodland, wet Vs dry!

Edit: even by reverse engineering from typical species solely associated with a certain habitat. I think AI should start showing us exactly how intelligent it is!

But I suspect inaturalist only relies on other sightings from the same area.
So if you are running a moth trap in the only small patch of coniferous woodland, on private land never visited before, surrounded for miles by broad-leafed woodland, you might get less accurate suggestions .

Probably - and I don't know how large an area it uses.

 

[Surreybirder]

In my, admittedly limited experience, iNaturalist will often get you to the right genus but rarely to the right species. But you'd have to do a lot more testing to evaluate it properly. (It's like these adverts for cream on the TV, 78% of 47 women agree that their skin feels younger - but what if you test it on 1,000 randomly selected women?!)

 

[pppppvvvv]

I can believe the inaturalist results would deteriorate for an experienced Moth'er, who only uses it for problematic micros say.

But your average (UK) punter, or "new moth'ers" putting all their photos into it, I'm pretty sure, well I can vouch for it myself, the results are pretty good, and records are verified by others.