Flag or otherwise mark taxon photos whose CID doesn't match taxon

Hmm I do think I’d like to be able to keep the option of using a photo that doesn’t have a CID matching the taxon. Particularly for subspecies because they don’t move the CID forward. I fairly frequently want to add a subspecies photo and the CID is at species.

3 Likes

I agree with the initial request, it would help.

I suggest a red bounding box around each picture whose source observation has a current ID not in the scope defined by the taxon page ID.

I wouldn’t suggest to remove anything automatically from the set of preferred pictures, because:

  • marking would be enough to satisfy the need,
  • removing automatically looks good, but if the current ID changes back, the picture won’t come back in the set of preferred pictures, which would be worse.
4 Likes

I’m a relatively new user of iNaturalist but twice I have come across a taxon for which the first default or reference photo appears (to me) to be incorrectly identified. When I view the observation the image is part of, sure enough the Community ID has been updated to another taxon at Research Grade with no dissenting IDs, and yet the photo still appears as the default for the previous ID. Once I figured out I could remove that photo and put a new one in its place, I did that, but this seems like something that would perpetuate mis-identifications throughout the database. Why not add a flag or pop-up message alerting the user that this particular image has been moved to another taxon? Maybe a red “!” that, when you click on it or mouse over it, explains the issue. Or maybe a curator gets an alert to remove the photo from its default position. Or perhaps simpler: once it gets a new Research Grade Community ID, the image would be automatically removed from the default photo group for the previous incorrect taxon.

3 Likes

Thanks @suecar! I moved your post over to the existing topic.

Thanks @bouteloua! I tried searching but didn’t find this previous thread (obviously).

Came here to add this feature suggestion and (fortunately) discovered that it had already been proposed. It’s great that iNat has taxon photos, and I do understand that earlier in the project’s evolution there was a need to find photos for a whole lot of taxa when there weren’t yet a lot of well-curated RG observations. This resulted in taxon images being added based on observations whose community ID is now not a match. Similarly, a lot of images were added from Flickr (and to a lesser extent EOL and Wikimedia Commons).

It would be wrong to automatically purge any “misfit” images, but it would be great to highlight them. As suggested by @jeanphilippeb and @suecar, a red bounding box or exclamation mark would seem appropriate, maybe with a popup note along the lines of this:

Although this image was chosen to illustrate the species Eleustrine latifolia, the observation it was taken from currently has a community ID of Cobana campanulata. You may want to consider replacing this image with one from a Research Grade iNaturalist observation of Eleustrine latifolia.

Given that the need for off-domain image sources is now much reduced, I think it would also be appropriate to have a similar yellow warning for non-iNat sourced images:

This image was chosen from [Flickr/EOL/Wikimedia Commons] to illustrate the species Eleustrine latifolia. We can’t be sure whether the image creator applied the correct identification and you may want to consider replacing this image with one from a Research Grade iNaturalist observation of this species.

8 Likes

Please don’t make new requests in a topic already dedicated to one request. If you would like to discuss a related topic, you can create a new linked topic. If you are ready to submit a request, please formally do so.

1 Like

I agree with @brian_d that it’s useful for people to discuss alternative solutions to address the problem that a feature request is intended to solve. If the proposer of the alternative or another supporter feels strongly enough to flesh out the details in a separate feature request that’s great. But we need to be able to talk through alternative options.

In this particular case, I feel that automated highlighting of mismatched taxon photos would have a more positive effect than adding an additional approval step.

I’m not aware of a major problem with non-curator users adding inappropriate photos to existing taxa, although I can imagine this happens for Felis catus, etc. Therefore, it doesn’t seem that making curators solely responsible for fixing old, incorrect photos would be very effective. I think this would cause an unnecessary bottleneck.

4 Likes

I was about to post this, so I’m putting it here:

I (also) suggest that when the ID for an observation changes we ‘flag’ any taxon pages containing images from that observation for a curator to see if it should be changed.

I regularly check the details of the images on taxon pages and I often find the photos are of a different species. Yesterday I found Eriophyes sorbi has only one taxon photo, which turned out to have be misidentified 3 years ago and a year or so later got a new ID. So this taxon has an image of the wrong species for about 2 years.

I also found that Canadian Bunchberry had an image of a Western Bunchberry for years.

It’s mostly an issue with more obscure species but it’s kind of eveywhere when you start to check.

5 Likes

I see this fairly frequently. Another example from today:

https://www.inaturalist.org/taxa/278748-Phoradendron-dipterum

The photo was initially Research Grade, but has subsequently been re-identified, and sits at genus level for now. Meanwhile there are other Research Grade photos of the species that have not been added to the taxon page. It would be great to have some visual indicator of the fact that the photo is no longer considered to the be species it is supposedly representing.

2 Likes

I suspect that we would have to run some sort of process in the background, checking for mismatches. For example, right now we run a process every hour or so to see which accounts have met the requirements to send messages or make places/traditional projects.

I don’t know exactly how much it would affect overall site speed, but having functionality of this kind would probably a trade-off of some kind.

2 Likes

Maybe it could be implemented as an on-demand check-up for a single taxon’s taxon photos (within the edit photos modal), instead of as a constant global process?

5 Likes

I’d support that idea. Another option would be a process that ran once a week or month - that would still be very useful in recognising these.

If necessary, the automatic weekly/monthly check could be restricted only to taxa with, say, fewer than 100 observations (similar cut-off for what gets included in Computer Vision) as those are the most likely to have an erroneous image.

3 Likes

Another example highlighted here: https://forum.inaturalist.org/t/cover-image-photos-that-appear-to-be-incorrect-or-dont-display-any-feature-to-prove-the-id/33242

Another, perhaps more elegant, way to deal with this might be to identify any observation when its photo gets used as a taxon photo (source_of_taxon_photo=true) and then run a check only for those observations, and only whenever the Community ID of the observation changes. In such an event, that would trigger a process that would flag or mark the relevant photo on the taxon page. I don’t know enough about programming to know whether that would be a more efficient way of doing this.

4 Likes

Or if iNat would edit taxon pictures sourced from Flickr.
Any photo which doesn’t link to an iNat obs.

1 Like

I think it’s fine for this process to be implemented in whatever way is most efficient.

If the check happened automatically (as opposed to on a background schedule or on demand), I assume that it would need to be triggered as part of an observation receiving a new ID. The logic might be something along these lines:

  1. Has the Community ID changed for this observation?
  2. Are any of the photos from this observation used as taxon photos? (This implies a lookup table to find taxon IDs from a given image ID.)
  3. For each photo, is the taxon the same (or an ancestor) of the CID taxon? If not, set a flag to note the discrepancy.

Depending on your existing data structures and indexes that process might be very quick or quite processor intensive. In the latter case, it might make more sense to have a scheduled background process that loops through all designated taxon photos for all taxa and checks the current CID.

I’d prefer to avoid doing this on demand if we can. If that option is chosen, then it will only get used by people who are aware of the issue.

2 Likes

Three more examples of this problem from this thread - further evidence that this feature request would be helpful:

2 Likes

That’s fine, as a filter, because extremely few observations would have this status, in comparison with the whole set of observations. As a consequence, no need for a background process.

1 Like

Another example of a taxon photo that does not match the ID of the source observation. Any update on the suggestions made to address this issue, @tiwane?

3 Likes

I also think this is a good idea. It would need to be run once on every observation that provides a taxon photo, but thereafter it would only need to be triggered when a taxon photo is changed, or when the CID of a taxon-photo-providing observation changes (not every time an ID is added). A flag seems the best approach because it would be searchable.

2 Likes