49 people clicked the “like” button on your “end of the Midgepocalypse” post.
A debate about “should we do A or B?” will always get more replies than “I’m doing this cool thing.”
I appreciate your work on chironomid identification and CV correction, and I hope you keep it up, no matter what happens with the default photos.
On topic:
It does seem to me that the ideal way to handle default taxon photos would be images that are edited to show both the whole chironomid and the genitals. Or both the caterpillar and the butterfly/moth. Or both the male and the female.
But someone would have to do the work to build all these images outside iNat, and import them. (with rare exceptions, like my favorite default taxon photo, the northern cardinal)
Also compare the default photos that iNat staff made for mammals, chordates, and all other taxa that include humans.
Here be dragons.
Image editing, if becoming tolerated on iNaturalist beyond (unavoidable) exposure/crop/white balance, shall be made abundantly evident;
and
Heavily-edited images should perhaps be excluded from “Computer Vision” training (if there’s a way to do so… marking Casual?).
Or else… unsuspecting newcomers, or maybe the naive Computer Vision algorithm, may end up thinking that Argiope sector is a spider with a red circle on the torso and a big red arrow-like organ sticking out (ok, too extreme an example here :-D). Or is that a fungal disease? spider ring-measles maybe?
Slightly off-topic:
– In some cases I wish there was no taxon picture possible at all - wherever photo(s) are notoriously treacherous and unhelpful for ID, because thousands of taxa look more or less the same while lacking any uniquely distinctive feature to be circled in red. Thinking of you hawkweeds and dandelions. Taxon pics can eventually do more harm than good there.
– Knowing how IDing-by-photo is frequently error-prone, I personally treat taxon pics as ‘visual memory cues’ and nothing more: little colorful icons that spare me reading some taxon name; whenever the image changes I’m a bit lost at first.
Too difficult to find a guideline. The first Dipteran photo might be showing wing venation good enough even if it is not the wing alone; etc
and some species require other details such as frontal angle, setae/bristles on legs etc; a good identifier may know all this for a species, and can accordingly choose photos and than hope that nobody will change again and make it worse.
I agree that edited images with the full body shot and an up-close photo of the genitalia would be a best-case scenario for the taxon images. But, for reasons outlines above, I do not believe just the genitalia work as effective default taxon photos, and a good majority seem to agree (over two-thirds of voters).
I just had a thought that clarified for me what I think ought to be done, as I also would say that a photograph of a whole-ass oak tree is utterly useless.
The taxon photograph should be as close to a pressed/pinned/mounted/taxidermied organism as possible; showing all features.
In the case of a plant, this would be a photograph that resembles that of a pressed herbarium specimen; with top and bottoms of leaves visible, leaf layout, flower structure, bark. All that stuff, or as much as possible visible in a single photograph.
In the case of an insect that’s a full-body organism that’s been pinned, so the photograph should show that in the highest possible resolution.
For a mammal or fish or whatever that’s probably going to be a full-body shot showing a side-on photograph.
If I were to add a newly discovered frog to a collection, it would be a full specimen. A newly discovered plant would be a pressed specimen. A bug would be pinned to a board. A bacteria would be colonies on a petri dish.
No, thank you. I want my plants ‘green’ alive and growing - as I might see them while hiking.
Assessing pressed herbarium specimens is a different skill set. iNat sets out to encourage ‘normal’ non-scientists to engage with nature.
Idle curiosity. How would you press something like a king protea flower? Some brown woody bits, which in life can range all the way from pure white to cherry red.
For your oak tree. A leaf, an acorn, bark? and yes a wide view of a whole tree.
Pinned insects have weird / dead colours and crucified anatomy. Include that in the taxon pictures for entomologists, but show us the live colours and anatomy too please.
Most species cannot be adequately represented by a single photograph. This is especially true for organisms with radically different life stages or significant sexual dimorphism. Specimen type photos are very useful, but in most cases not for the taxon photo, which has to work as a thumbnail and also portray the organism as it is usually seen by observers in the field. The taxon photo has to be backed up by other photos which show specific details appropriate for the taxon.
Even in a really simple case like the Panama Pearl Oyster, multiple photos are needed. What is the best taxon photo - one that shows the living animal in situ, one that shows the pearly interior or one that shows the leafy periostracum?
This is why I completely disagree with the idea regarding pinned insects. Moths are pinned with the wings fully spread to show all the features, but a live moth that habitually perches with wings folded will not resemble the pinned specimen.
A big plus one for genitalia as taxon image on complex taxa like this!
And wholehearted support to you @zoology123 for dealing with this family and fighting the good fight as well as arguing the case for this here!
Sarcophaga carnaria ( and other members of genus ) were consistently incorrectly chosen through autosuggest. We changed them to genitalia and it’s much better I think - almost all members of the genus are unidentifiable without dissection…it serves no purpose to have the full body as the default image - it only increases the likelihood of naive comparison by newer users using compare tool or autosuggest.
I say that as both an identifier of Diptera but also someone who uses the autosuggest myself as a sort of placeholder when uploading at times…especially in taxa I don’t know. Seeing genitalia as a taxon image ( or for example, cellular detail in a moss taxon ) would be an immediate indicator for me to take care.
@raymie - I think the voting by random people on the forum is meaningless unless those voting are all familiar with the issues visible in identification in Diptera ( or similarly complex and under-observed taxa ). Often in these conversations in the forum, the people present and vocal seem to be more active in bird ID and plant ID… where on the whole, it’s simply an incomparable area. More respectful here I think would be to take that vote within a flag tagging in the key identifiers of the family. Ultimately they will see the issues better than anyone…and imo should be the ones with the say in curating the taxon photos ideally.
In this instance, in any case, Chironomidae are overwhelmingly identified by @zoology123 so imo they should absolutely have the say in what makes sense for the default taxon pics.
Please don’t do anything to deter their fine work!
We’ve updated the Computer Vision Suggestions on the web to display Taxon Photos that are most visually similar to the observation photo.
…
Previously, the Computer Vision system displayed the same taxon photo for a species regardless of the observation. Now, it selects the most visually similar Taxon Photo for each suggestion.
Also, the taxon photo guidelines were modified on the same day as that blog post (which links to them). I don’t know what changed, but make sure you’ve read the current version.
This whole conversation is largely irrelevant now. I have had to adapt to the changes. Like most changes there are positives and negatives. For this change there are much more overall benifits than negatives, and while genitalia wont be the first appearance now ever for the CV unless the image is of genitalia. There are ways to compensate and still try to reduce misidentifications.
Mainly by filling taxa with as many quality photos as you can. In a sense many of the images of reccomended taxa will just look visually the same now. Which is arguably good or bad.
Also under the taxa photo guidelines.
“Taxon photos are the first images people see when they research a taxon, so photos that are helpful for identification should be prioritized.”
“Photos should be clear and show diagnostic features of the taxon.”
If the diagnostic features of a taxon is mainly genitalia, then this supports genitalia being the first image rather than a full adult image that lacks the details. Even if the photo update didn’t happen, these guidelines favor what i argued above. Images for identification.