Why are there these new “rules” regarding edited photos? I am in full agreement with the AI-generated flag but these new guidelines regarding edited photos are extremely frustrating. For me, a DQA downvote on an observation I don’t downvote is frustrating for me.
“Removing or replacing elements from the organism or the scene “- If I want to remove a person in the background since their face is visible, why is that an issue? I understand the organism, but the scene doesn’t make sense to me. I interpret the Inaturalist guidelines to mean “if it is identifiable, not planted, geospatially accurate, and appropriate…then it meets RG criteria”.
“Adjustments to: contrast, color, saturation, shadows and highlights, white balance, lens distortion and vignetting, sharpness etc., that fall outside of general accuracy (e.g. making the subject much more saturated or vibrant than it naturally appeared) “-IMO as long as this doesn’t change the identification i.e. making a Canada goose completely green, this should be fine. I could lose thousands of observations to the endless sea of Casual grade observations if this was enforced rigorously, and even many OOTD’s are enhanced by camera or computer software! In many cases, changing the brightness, shadows, highlights, sharpness, etc. has helped lead to the correct identification since the original photo was too blurry or distorted from glare or darkness.
“Sketches or drawings made from memory”-I have an amazing memory and if the organism is drawn properly should it then be made captive? Up until today sketches were evidence of organism, and now they are only evidence if you are sketching in the field? I’m sure a lot of other Inatters could remember how to draw a Canada Goose from memory that they saw a few hours before.(I sure can)
These new “rules” seem to bring more harm than good in my opinion, and if you have thoughts on this please share!
This is really a classic case of “a few bad actors are why the rest of us can’t have nice things”. Or at least why the rest of us have to muddle through extra complications.
These changes are a good faith effort to deal with instances in which people use technology to change or fake photos for whatever wrong-headed motivation drives people like that. The point isn’t to arbitrarily make it harder for users who are here for the right reasons. We just have to be patient with some guardrails in place to keep things from descending into chaos. A lot of life is like that… traffic lights. Locks on doors. Passwords. It often feels obstructive. But remembering the motivation behind these restrictions makes it less annoying.
The iNat staff are doing their damndest to keep this place nice, and they’re doing a pretty great job of it. Knowing why they’ve implemented these changes helps guide you as you decide whether or not some alteration is appropriate. :-)
This has always been the case. Human memory is fallible, and drawing what was seen later after-the-fact has never been acceptable as “evidence of organism”. If we allowed everyone who declares “I know what I saw!!” to draw their perception of what they encountered and count that as evidence, we’d literally have to add bigfoot, aliens, Chupacabra, and the Fresno Nightcrawler to iNat’s taxonomy. After all, perfectly intelligent people who think highly of their own memories will swear that they saw them. And if human memory is that bad, then I sure as heck don’t trust anyone, myself included, to remember which species of woodpecker they saw a month ago if they didn’t record it at the time. I have a large moth collection, and I can’t count the number of times I’ve thought “oh yeah, I caught one of species Y in X county” only to go look at the specimen and find that I’d misremembered. Having the humility to accept that none of us remember things nearly as well as we’d like to believe we do is an important part of doing science. If you sketch what you’re looking at as you’re looking at it, that’s much more reliable that recalling what you saw later. Every time we recall a memory, we alter it as we recall it, and memories become less and less reminiscent of what really happened every time we’re forced to re-tell the story. That’s not a judgement on anyone’s intelligence; it’s just human psychology.
All of these were taken on my camera and had some minor changes to brightness, sharpness, etc. on each. All are still verifiable and RG.
Should these and many others go straight to the Casual dump because there were some minor changes? I have thousands of camera photos that meet these criterias, but these five are some that I got super close up and clear(and are some of my favorite photos on Inat.)
I can change in the future but the past has been done already. This would make me extremely upset if these observations were made Casual even though they are completely accurate.
Under “Accepted” it clearly states that the following is perfectly fine:
Standard adjustments to: contrast, color, saturation, shadows and highlights, white balance, lens distortion and vignetting, sharpness, noise reduction, etc.
Under Not Acceptable the key aspect is the following:
*that fall outside of general accuracy*
Obviously different people will have different ideas on how far the former can be pushed until it falls into the latter, but the idea seems clear enough that edits that don’t drastically change the image are ok.
The way I interpret it, minor changes are perfectly fine. “Standard adjustments to: contrast, color, saturation, shadows and highlights, white balance, lens distortion and vignetting, sharpness, noise reduction, etc.” are specifically listed as acceptable, and so are focus stacking, HDR, collages and annotations. I would add to this that probably the use of lens filters (e.g. reducing reflections and popping up color saturation with a circular polarizer while taking the picture) would also be fine. It’s “adjustments … that fall outside of general accuracy” that are an issue. Of course there’s going to be some variation in how this gets interpreted and where people draw the line between “standard” and “falling outside” the acceptable.
Just for sake of discussion, here’s unedited (left) and edited (right) pictures of tulip tree blossoms in the fog. Would the edited version be considered too “fake” for iNat or would this still be acceptable?
As an observer, I think one approach if posting a heavily edited version might be to also include an unedited version and a note explaining what kind of editing was done and why (e.g. in the example above to visually remove the fog). That way the original evidence is still there and identifiers can choose to rely on that. As an identifier, I would appreciate that. I think as long as there is also an unedited version of the evidence available, the observation as a whole should not be subject to this new DQA mark. If the editing is done purely for artistic reason and is the only image posted, maybe a website for artists would be better suited than iNat.
For me I don’t actually know how to add multiple photos to the same observation when batch uploading from a computer. It’s why my bird photos are single photo only.
Just blur out the face, that would be fine. But using something like generative fill (like the example I shared in the blog post) adds entirely made-up elements to the scene, which is quite different.
Yes, this is what is meant. I’ve seen hundreds of thousands of photos on iNat and there are only a handul I’d mark as “no” for this DQA. Standard editing is fine. But for example I’ve seen plenty of photos shared online (not on iNat) of San Francisco Garter Snakes that have had their saturation and vibrancy jacked up to a level well beyond anything you’d see in real life.
I spent a couple of years as an art student, have an excellent visual memory, and can produce drawings (of certain taxa) that may be more accurate than most photos taken by the average user. I’ve actually thought hard about whether to include field sketches in iNat observations. I fianlly decided not to, for two main reasons: (1) If you don’t know me and don’t know my work you wouldn’t be able to assess how accurate my drawings are…I decided no one should have to accept my drawings on faith. (2) Taking photos is a lot faster.
I also feel that in a project with as many users as iNat has, if the data is going to be scientifically useful, you have to set strict definitions as to what constitutes evidence. Yes, a trained field biologist, or a trained artist, can produce drawings that constitute valuable scientific evidence – it sounds like you are one of those people. But you have to know that individual’s qualifications, and in a mass project like iNat, the scientists who use our data can’t know everyone’s individual qualifications.
Having said all this, hope someday I meet you in the field with our sketchbooks in hand!
As others have gently pointed out, I think you are reading the new guidelines on photo adjustments too harshly. If any adjustments (cropping, rotating, overall brightening, reducing or enhancing contrast slightly) are aimed at, and accomplish, a more readily identified organism, that’s certainly within bounds. I would have to say probably 99% of my images are post-processed to some small to moderate degree. In the raw images, this mountain-tanager in Colombia was just a small distant dot on a tree-top on a dark drizzly day. Only by editing those images rather heavily (cropping, and altering brightness and shadows) was I able to make the bird recognizable.
By contrast, I have encountered a few examples over the years (very few, in fact) where an OP has increased the “vividness” (e.g. contrast and saturation) of some normally dull gray-brown moth to such an extent that it bears no resemblance to the real world creature. That should be viewed as unacceptible alteration. I have cautioned a few iNatters about using any automatic “Vivid” settings on their phone camera or small point-and-shoot cameras because such a setting commonly distorts reality to an unacceptible degree.
The way I’ve thought about this is less as a “bad actors” thing, because there haven’t been that many examples of this on iNat.
I’ve been thinking about it more broadly because these tools are out there and they are getting normalized, and their usage is just going to grow. I suspect in a lot of wildlife photography communities, tools like generative fill are probably pretty widely used and accepted because in many ways it’s more of an artistic endeavor than a scientific one, in that context. And people maybe don’t want that blade of grass over the lizard’s eye.
So the goal is to make it clear that, regardless of what you post to Instagram, Flickr (hey, I still use Flickr!), etc, those manipulations are not for iNat. On iNat, photos are evidence and should accurately depict your encounter, and we want to try and get ahead of the normalization of these tools on iNat. Also, to allow people to vote on evidence accuracy the way they can on other aspects of an observation.
In many ways I don’t think this is too different from the copyright infringement flag. Taking other’s content and posting it without attribution is commonplace all over the internet, so I think some people (with no malicious intent) do that on iNat becuase they don’t realize what kind of photo is acceptable for an observation. And we have ways to address that on iNat. This is in a similar vein.
Batch uploading may be one thing.
Automated severe picture enhancements in batch mode are for sure another thing.
Please understand that for the step of taxonomic identification, there is no process available for batch identification.
The same, as it is expected from you to click yourself through the initial ID step for each of your photos while applying due diligence, others need to do the same and understand what is shown on each individual picture and for what it gives evidence when reviewing your content.
I hope you take the same effort at your end than what you expect others to apply to your observations. I would not see how automated photo enhancements could automatically ameliorate the evidence for observations in each individual picture. Some will show up bad pending how settings were chosen. Given that this is then frustrating to a multitude of reviewers, i believe it is better for people, process and data quality to bring the frustration back to the originator (reporter) right from the beginning.
This statement written by someone who tries to limt himself to three pictures max per observation, selecting those from a multitude (if the object allowed) by relevance for to ID (requires initial investigation) and for general picture quality.
I understand the need to clean or enhance (not modify) a photo for easier identification. Lighten; Brighten, etc.
I also understand the need to prevent the potential creation of AI-generated organisms.
My suggestion: Add the original as a 2nd photo. This way, any person can then see the original and know that the manipulation is purely for clarity purposes.
Further, if necessary, add a comment explaining the reason for the enhancement. (E.g., Leaf enhanced to highlight venation.)
I would like it if any of my observations were downgraded to Casual; I get a notification regarding this. This way, I can defend my decision and not discover it months later by chance.
I hope that users interpret this guideline in the spirit in which it was intended, but I want to mention another consideration:
I have a couple of lenses that produce fairly substantial optical effects under certain conditions (ultra-wide angle; mirror reflex), meaning that images taken with these lenses may not reflect the scene as it would normally appear to an observer – not because of any editing, but because this is how they came out of the camera. And in fact it would require more editing skill than I possess to make these photos appear normal.
Interestingly, it appears I have not uploaded any of the more serious offenders as iNat observations, perhaps because I felt that they were too artistic or too weird, but I hope if I were to do so they would not be subject to flagging.
The mirror reflex lens in particular seems relevant, because it is something people are likely to turn to not for artistic purposes, but in search of a compact and lightweight telephoto (though, like me, they may conclude that they are often difficult to use for the subjects an iNatter is likely to want to capture with a long tele).
I totally agree.
A photo isn’t and never was showing “reality”.
It is a social compromize for showing evidence of something.
It’s 2 dimensional and can be manipulated by physical, chemical and now also by several levels of electronic means and nothing is very transparent to the observer.
Like for every social compromize, guidelines are required.
And as long as originators clearly comply to the scope of INaturalist and do no harm to people or data or process, it should be fine.
If not, not the guideline is to change. Push it back to the originator of the problem. Who else shall adapt?
Just wanted to chime in to say that I wholeheartedly agree with @paul_dennehy and others above in support for the explicit requirement that field sketches need to be made from life (not memory) to be counted as evidence. There’s all kinds of research (pop sci example) showing that humans are vastly overconfident in their memory. I personally have a very good memory, but I am still overconfident in it and make associated mistakes.
Human memories, for almost all people, are not complete records like a picture, but assemblages of fragments of details and previous memories recombined by the brain. In the case of remembering what we saw in relation to an observation of something like a squirrel, the brains of almost everyone will not have anything analogous to a photo/video of a squirrel stored away and replayed for them. Instead, they will remember something like “I saw a squirrel with these features move this way under these conditions” and then their brain will recreate that scene when called upon. In fact, each time we call up a memory, we recreate/reimagine it, and the memory changes. Remembering is a dynamic process, and not passive recall. It is actually not totally unlike how an AI model (ie, some type of artificial neural network) creates an image from a prompt by accessing related “memories” (ie, photos, etc.) in the AI model’s training data (analogous to humans’ previous experiences). In short, if a human thinks, “I’m going to draw that squirrel I saw earlier today” they will end up most likely drawing a squirrel, not because it is what they actually saw (though that is likely to be the case), but because they have already passed judgment on the memory, categorized it, and labeled it and are recreating it according to their previous perception of what they saw.
Drawing from life is a different process, as one can record specific details from the individual and create a more detailed, true-to-life record that is much less filtered/transformed by the brain. On a side note, this is one of the fundamental reasons why many artists go into the field to make art or bring specific objects into the studio to draw, paint, etc. They could, of course, just create art from memory without the hassle, but this process often leads to the creation of art that lacks the details, vibrancy, and a feeling of “reality” that a work made from life will have.
For what it’s worth, eBird has had similar editing requirements for uploaded photos ever since they started allowing checklist media (although it’s harder for the rules to be enforced there): eBird/Macaulay Library Photo Upload Guidelines
Exactly this. And if anything, a talented artist who has drawn/painted 100 gorgeous life-like squirrels is even more likely to add details that they don’t specifically remember from “today’s” squirrel, because they have a very detailed pre-conception of what “squirrels” look like. Just because you can create a beautiful rendition of a squirrel doesn’t mean your painting of the squirrel you saw last week will capture that squirrel in particular. Especially if you’ve misidentified something (which we all do all the time), the rendition you produce will show the species you misidentified it as, because that’s what you remember seeing. No one who misidentified a Lesser Scaup as a Greater Scaup is going to later draw a depiction showing Lesser Scaup characteristics, because they “remember” it as a Greater. They might draw the most realistic Greater Scaup ever portrayed, and no one can argue against their ID based on the picture they produced. Camera lenses may be capable of distorting images, but they’ve got nothing on the distortion power of the human mind.
On the case of increasing saturation, many Erotylid beetles lose a lot of color after they die making it very hard to see the patterns which are often very helpful in IDing them (this observation is a good example). I’ve recently gotten two specimens of an Erotylid that had its pattern badly faded, one more so than the other. In the one with the less faded pattern, I increased the vibrance a decent amount (I noted this in the observation) to make the pattern resemble more what it would look like when alive and be more visible: https://www.inaturalist.org/observations/313996566. Would this observation be at risk of becoming casual since the vibrance was increased beyond what it appeared in real life, even though it doesn’t change the shape or general appearance of the pattern in any way other than color vibrance and making it easier to ID?
