Difficulties with Identifying

I understand that far more people are observing than identifying. I’d like to help with the latter, but have found it to be technically less easy than I expected.

For instance, on the app on the phone, the AI suggests a species which you can confirm or not. On the website, it does not do so.

Also, I’d like to add ID’s to observations marked “Unidentified.” It seems like it should be easy to see only “Unidentified” observations from a given place. But I’ve yet to figure it out.

All this probably because of my minimal expertise. But it seems like it could be easier for people like me to help address the imbalance of observers and identifiers.

Thanks so much!!

4 Likes

The website does offer Computer Vision (since its not really AI) suggestions. The boxes for the CV suggestions are in different spots than the app. You can go directly to an observation page, click the “Suggest an Identification box” than click “Species name” without typing anything in that box (if you type something in that box, the CV option doesn’t appear)

or you can go to the identify page (https://www.inaturalist.org/observations/identify) ((It can be hidden under the “More” tab depending on the size of the screen))


click on an observation in the panel

then click on “Suggestions”

once you click that panel, this screen will show up:

click on “Visually similar” (or else it won’t show Computer Vision suggestions)

then you will get CV suggestions

Just as a note: Computer Vision suggestions are just suggestions. They are not always correct, especially with Fungi, less charismatic species of Plants and Insects, etc. The CV is usually pretty great with Birds and large mammals though.

Also keep in mind that CV suggestions are biased towards the US/Canada, due to so many of iNaturalist’s observations coming from those two countries. This used to be a much larger issue, but its now starting to even out (but its still biased!).

13 Likes

Theres two ways of seeing unidentified observations on iNaturalist’s identify panel:

You can either add ?identified=false (or &identified=false if there are other filters in place) to a URL (examples: https://www.inaturalist.org/observations/identify?identified=false [This URL is global unidentified observations] or https://www.inaturalist.org/observations/identify?place_id=1&identified=false [this URL brings up unknown observations just from the United States])

or you can go to Filters → Iconic Taxa → Unknown

Using the Iconic Taxa filter will also bring up State of Matter Life and observations of various Bacteria, so I recommend using the ?identified=false URL.

10 Likes

It is very commendable that you want to start identifying. Just don’t trust the AI suggestions blindly. Even though INaturalists AI is quite good, it should always be double checked with other sources. Stick to the level of ID you are comfortable with.
If you want to identify “unknowns”, you have to choose that category in the filter options.

16 Likes

Welcome to the forums!

1 Like

Thanks. Glad to be here ;)

1 Like

Awesome!! Many thanks1

I also mostly rescue things marked as unknown. Here are my two sources of things to identify:

  1. select a date (about 1 year old) and select the options “unknown” (I keep location as global, since I almost do only coarse IDs, frogs in China look the same as frogs here and both just get marked as Anura) and a few others, e.g. https://www.inaturalist.org/observations/identify?quality_grade=needs_id%2Ccasual&page=93&iconic_taxa=unknown&photos=true&created_on=2023-06-06&createdDateType=exact

  2. when I find a very obvious organism (e.g. an insect) that was labelled “unknown”, I go to the observer and walk thru their observations instead. This list has a much higher percentage of recognizable organisms than the global list mentioned above (which ist mostly weeds and trees). This just gave me a few hundred Lepidoptera from one person over the last few days.

In both data sources, I start on the last page and move backwards to avoid missing things that slipped to a previous page by reviewing previous ones.

The reason for using an old date is that there are observers which produce a very high volume of observations and perform mass identification of their findings a few months later, so I don’t interfere with their process.

My process is as follows: If there are many identifiable things on the page, I walk them one by one using the keyboard navigation of the ID module. If there are only a few, I just click those. The other question is what to mark as “reviewed”, i.e. things I never want to see again. I use that on trees (I’ll probably never learn them), exotic weeds, blurry images and a few other things. When I’m walking the observations of one single user, I mark everything as reviewed unless I’m particularly interested in that (in which case I follow that observation).

I leave dicots unchanged (unless I know at least the family level or so) - moving them from “unknown” into “Magnoliopsida” would interfere with the workflow of at least one IDer and probably not help the others.

Note that IDing an observation marks it both as reviewed and followed.

Every few days I look into my notifications in order to see if I was wrong (in which case I withdraw my ID); or someone else was wrong (happened only once so far), then I add a comment. Then I unfollow the observation (unless it was the observer who confirmed my ID, some people do that without being sure).

From time to time I look into https://jumear.github.io/stirfry/iNatAPIv1_identifications.html?user_id=ralfmuschall&current=false (you have to insert your name in place of mine) to see disagreements I might have missed. Don’t click on links there (for some strange reason this logs you out of Inat), rightclick, select “copy link” and paste that link into a new browser tab.

6 Likes

This should have been phrased slightly differently: “Even though iNaturalist’s AI is quite good in high-volume areas for frequently observed species…” I regularly come across situations where it seems to be guessing wildly.

6 Likes

Hey there!

Totally get where you’re coming from - identifying species can be quite the puzzle at times.

As for adding IDs to those “Unidentified” observations, it should definitely be made simpler. It would be great to have an easier way to filter through those specific observations from a certain place.

It’s all about making the process user-friendly for everyone. Don’t worry about your expertise level – we all start somewhere. I’m sure with a bit more digging, you’ll uncover ways to navigate the platform more efficiently. Keep at it! Your efforts to bridge the gap between observers and identifiers are truly appreciated. Kudos to you for making a difference! :blush:

4 Likes

That’s what one expects from an AI. Young dicots look all the same, so many observations remain unidentified and the AI has no training material for the species. As a result, unknown images are projected onto the few species it knows (in my region foxglove, Lapsana and a free more).

I think more people should upload identified images (which probably would mean captive plants, as it is hard to know wild ones before the have flowers).

What would be a proper approach for observers to tackle this problem? The rules say if I come back later to the same plant, I shall create a fresh observation instead of adding the picture of the flower to those of the baby plant. As a result, the CV will never learn from my recently improved knowledge about this plant and not have training data for the young stage.

1 Like

Just upload the grown plant as a new observation, then put a link to that new observation in the notes of your baby plant observation and change the ID!

4 Likes

A new annotation for juvenile leaves ?
We have larva for insects, why not seedling for plants?

1 Like

I’m unable to find a larvae project. I often see unknown “worms” that I cannot put into beetles, butterflies or sawflies because the image doesn’t show the ventral side. “Stage of life” for pterygota (correctly) doesn’t offer larva, so I cannot even mark them in such a way that interested IDers find them easily. The proper solution to this would be having Endopterygota in the taxonomy, but I understand this is hard to do so we won’t get it anytime soon. A project “larvae of Endopterygota” might help for the time being, but I couldn’t find any, only those in the screenshot (the invisible one below the bottom is something with neuston, i.e. water).

Other question(s): am I correct in assuming that “Beach Blobs” is for marking unidentified things that then somebody IDs as cnidarians, red algae etc; but “Rosettes of Dicots” is exactly the other way around: only put things there which are known at species level to help others (and CV)?

If yes, can I usefully put unknown rosettes somewhere?

Is there a “Land Blobs” or similar where I can put unknown things that might be fungi, slime molds, uneaten prey intestines left by predators etc?

You could make a project and list what you want included.

Let’s go to the Identify Tab. In the box, key in ‘Animal’ . In the second box Key in your country. Alternatively, in the first box, you can key in Birds, Insecta, plants, Mushrooms and Fungi, reptiles, amphibians etc… The second box, can be a state, or a country.
Many unknown pictures are labelled within a time frame, sometimes I see it but yes I’m unable to search for it too. The uploader often have edited their uploads or identifiers have put a general label to it.
Try to identify at your country or your local areas in iNat. The flora and fauna are more familiar to you. However, I also try to test my skill level by attempting a wider area, firstly by venturing to a nearby country (in Inat) and then further away. Creatures get more difficult to Id as we travel further. Many plants and animals have a globalised presence after humans introduced those species to other places of the world. If you attempt to identify without entering the country , it is not as easy as it seems to get to species level. However, it is easier to put a label to it at some levels of the classification.
There is a need for reference sources. A trusted foolproof source will be excellent for minimising errors.
I found a trick but it is not foolproof. Duplicate another instance of iNat. Go to Explore Tab. Animal is too broad a category. Let’s try Mammals instead. Enter Mammals and your country. There is a Lime Green Box stating the number of observations. Click on the next box so that the species box turns Lime green.
By the way, the map-Grid-List option, I always set to Grid. Now, that is iNat’s list of mammals in your country. Why it is not foolproof is because a few rare animals may have not been identified or pictures may be wrong and yet nobody could say for sure that is wrong. so such pictures just languish there. It is a convenient way to copy and paste the name to the next instance of your browser’s tab. I use other reference sources too. There are creatures which are very well known that you know instantly what it is and creatures that you need to check up its name.
Yes, the AI don’t seems to suggest anything at times if you are identifying. The uploader will get the prompt from AI. As an identifier, sometimes I see the AI suggestions, sometimes I don’t.
Right click on the picture, go to Google Lens and see if Google Lens has some suggestions.

Done: https://www.inaturalist.org/projects/larvae-of-endopterygota

I’d love to get ideas for corrections or improvements.
And I’m not sure if it is recommended to remove animals from the project once they have been identified into a narrower group, maybe I could add that as a hint for the IDers who work with observations that are there.

1 Like

I don’t see why it would be necessary to remove observations that have been ID’d? People looking for larvae to help with can filter the project observations to show only those at a high level.

The ID’d observations might serve as a useful reference for people who remember having seen something similar before but can’t recall what it is.

However, at present you seem to have set it up as a collection project that includes all Pterygota observations regardless of life stage, so there is no way to add or remove observations or sort for larvae.

2 Likes

That’s what I just found out when trying to add https://www.inaturalist.org/observations/155300307 . I was intending to mimick the rules from Beach Blobs, but the website shows me (as the owner) a different view of the project settings than for Beach Blobs. What do I have to change to make it work?

If you edit the project, is there a note somewhere on the page with an option to change it to a traditional project? (I know a traditional project can be converted to a collection project, but I don’t know if the opposite is true.)

When creating projects, the option for traditional projects is hidden at the bottom of the page in a link; you have to ignore the two big buttons for collection projects and umbrella projects.