Identification of multiple pictures in one observation

How does iNaturalist identify pictures of observations?
If I take 2 or more pictures from the same animal/plant but from different angles and I am posting them separately iNaturalist identifies each of these pictures every time new ( like pictures from a Bivalve from the outside and one from the outside or a mushroom from the top=the cap and one from the bottom=gills) But how does the suggested identification works if I combine these pictures as one observation? Does only the first picture receive a suggested identification or all of them?
Thanks for answering this question

2 Likes

Only the first, I believe.

3 Likes

The Android App allows you to scroll through multiple photos, each of which gets run through the CV suggestions. I don’t think this can be done in the website interface, unfortunately.

3 Likes

This might be a problem because often you can’t identify a mushroom with just one picture from the top as you have to see the bottom as well to make a positive identification

Are you asking about the CV or just observations with multiple pictures in general?

If you upload multiple pictures to one observation, the CV only uses the first picture to make a recommendation. If you want to see CV recommendations for all pictures of a single individual on the website, you can just check each individually in the uploader screen before dragging and dropping them together into one observation. This can be helpful sometimes.

If you’re talking about observations with multiple pics in general, these are great, and should ideally have pictures of one individual. Identifiers would look through all of them and use the information to make an ID (ideally).

10 Likes

yes that what I am doing: posting different pictures from the same mushroom/bivalve and see what suggestions are coming up for the same object and decide which suggestion makes the most sense.
Thanks for your response.

2 Likes

Bear in mind that the CV isn’t meant to be used in isolation. It’s best practice to do your identification using outside resources (keys, field guides, websites, whatever). The CV can give you some pretty off-the-wall suggestions sometimes, especially for taxa with few observations.

8 Likes

It’s very good to post multiple photos for one observation. The Computer Vision will guess using only the first photo, but we identifiers usually use all of them. We like seeing the same organism from many angles.

7 Likes

Yes, multiple angles help immensely with identification.

3 Likes

There is also this tweak which I haven’t used

https://www.inaturalist.org/computer_vision_demo

From the associated Read More. Wow! Can’t hardly believe it - so it is only the difficult half identifiers have to battle with.

half of all observations identified in the first 2 days

1 Like

Correctly identified? :grinning:

3 Likes

Nothing is carved in stone. Even the ‘it was correct last year before they upended the taxonomy’ changes with new publications.

In my experience for NZ spiders it’s rarely correct. Clearly other taxa have better luck.

It is more interesting, when people who can / know discuss IDs.
I don’t see a lot that are wrong.

But MANY trapped at a higher level because no one can / will take it further. I am haunted by these … 41K and counting

PS today 41 - 2 = 39 thank you to invisible identifiers!

That one can really be infuriating. Yesterday I came upon what had been a Research Grade Royal Tern, bumped back to genus because the African population was split off as a new species. So I had to re-identify it again as a Royal Tern.

There has got to be a better way to handle taxon splits.

2 Likes

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.