Quality Issues rejection

[Svetlana28639737zwpi]

Hi.

Just got this batch rejected for "Quality Issues"

I don't mass upload. This was a batch of 5 images, all rejected.

Can anyone suggest what where the "Quality Issues"?

[Svetlana28639737zwpi]

Forgot to add. These are AI generated with SD. Only upscale during the generation, so preaty much no "after the fact upscale" issues. Used SDXL models that has large "native" resolutions.

[Jill_C]

I'm not sure what caught the Moderators eye in the first image, but it's quite soft overall.  Many of the windows and doors in the second image are not well drawn - lines are not parallel, incomplete, etc.   

[Svetlana28639737zwpi]

"but it's quite soft overall" Well that might be the problem. The whole point of the "out of focus areas" that you get with wide aparature is to be "soft".

The only (logical) explanation is that in case their alghoritam checks the images, it sees the image as "missed focus".

Otherwise a real person (reviewer) could make that distinction. At least the ones that have used the camera with wide aparature lenses.

Come to think of it, i did get a rejection on the same issue with some backgrounds i did. They had very heavy "bokeh".

No other apparent flaws.

As for the windows. They are by AI image generations standard "perfect".

[daniellei4510]

"As for the windows. They are by AI image generations standard "perfect".

AI assets are not judged according to AI standards. If you are attempting to create AI generated images that are intended to be photorealistic, they will be judged according to the standards of photography. The windows were the primary reason for that particular image to be rejected. Period.

There us no algorhythm checking images. It is done by humans. 

I think it's safe to say, the moderators are trained to understand depth of field. In the snow image, I'd remove the "snow" since it isn't heavy enough to look like it's really snowing and comes across to me as specks of dust. If I were to buy this image, the first thing I would was remove the specks of "snow." I'm not saying it's a bad image because of the snow. I'm saying it was rejected because of the snow. It's also a tad on a blue side, and since the sun is illuminating the snow, I would expect to see a slight yellow cast. 

[Svetlana28639737zwpi]

As for the snow. Nothing needs to be removed because those particles in the air only ads to the image.
There are "clean" images in that batch that got rejected (will upload one more), so no, the snow was not the problem.

And no, the images are not reviewed by real people. At least the AI ones.

In the past half an hour i heve analized the rejection pattern.

The only conclusion is that they are reviewed by algoritam.

Every rejection with images with heavy bokeh is when there is no distinct subject together with the shalow depth of field. Like 3d backrounds or landscapes.
If they are in full focus, they get accepted no problem. Distinct subject or not.
I will upload here the whole batch of the last backgrounds that got rejected, to illustrate the point.
There are 3d ai backgrounds accepted today that are in full focus but with a lot of mistakes in generation.
So the other poster argument "Adobe has definitely raised the bar for quality lately, and it's no longer a matter of "anything goes because it's what you get from Gen AI" simply doesnt't stand. MJ6 images gets acepted with large ammount of unnatural shapes and artifacts daily.

As for the "real reviewers". Not long ago the wait period for a review of AI images was almost two months.
Now is cuple of days. With who knows how many times more ai images uploads daily.
Wonder how that happend 😉
Maybe adobe hired hundreds of thousands new reviewers 😉

[Jill_C]

Regarding your comment "They are by AI image generations standard "perfect" - there is more and more evidence lately that Adobe is no longer satisfied with this level of AI generation "quality". We are seeing architectural images with these types of flaws being rejected regularly. Just scan back and read other recent posts in this forum for examples. Adobe has definitely raised the bar for quality lately, and it's no longer a matter of "anything goes because it's what you get from Gen AI".

[Svetlana28639737zwpi]

Have you seen the enlarged image i have posted? Is there something wrong with lines, frames? Are they not paralel, incomplete, etc?

[Jill_C]

The top of the door is not parallel with the frame.

[Svetlana28639737zwpi]

Any other flaws? You wrote "Many of the windows and doors"

So you are basicaly saying that the reviewer blow up the image to 300% to see that? 🙂

They do that with every image? 🙂 Are you serious? 🙂

[daniellei4510]

Probably depends on the moderator. But I suspect they all do at least 200% minimum.

[Svetlana28639737zwpi]

Have you look at the latest additoins to the library? Just sort the latest ai images by "most recent" and nitpick them like my images here. 90% of them would never got acepted. At a glance there are images of christmas gifts with ribbons fused with the box. Close up image of the woman with distorted eyes. Some would say that the top of the door that is 0.1 percent of the image is more important than the eyes on a portrait 3:)

There is a portrait of a child with 6 fingers and so on. On every other image.

All that approwed and added today.

So no. The windows and the particles in the air are not the problem.

 

[davide29060311]

I completely agree.

Sometimes one wonders if it's not a bot that is conducting the evaluations, given what some manage to have approved. Adobe has not consistently raised the quality standard lately for everyone. Either the validator was completely absent and approved a batch without checking them, or certain contributors have connections and manage to have anything and everything validated. Even a person with 2/10 vision in each eye could spot the issues in these images.

For example:
File Number: 672852943 - Published on December 6, 2023
File Number: 672818409 - December 6, 2023
File Number: 672371991 - December 5, 2023
File Number: 626346279 - July 29, 2023, but seriously?

I wonder how he manages to get them through!!

[Jill_C]

Ugh! Hands and fingers are the first place we look when trying to find AI flaws, and these are quite obvious errors. Would have taken me less than 5 seconds to spot them and push the reject button... It does seem that the Moderator was "phoning it in" on those images. Automatic approval for certain Contributors is not something I've heard mentioned before, but I suppose it's possible that certain Contributors are fast-tracked in the review process. I just found another image in that portfolio with too many fingers... 

[daniellei4510]

Yep. These are the accounts that give AI a bad name for sure.

[Svetlana28639737zwpi]

@daniellei4510 MP Studio is an acount that give AI a bad name? 🙂

That account has mass of images on top pages (by downloads) in all cathegories.

So he is selling a ton of images, with all of those obvious flaws. Adobe loves selling tons of images 3:)

Surely he doesn't get his images checked by the same reviewer 😉

I mean, judging by all of the flawed images he gets accepted 😉

It would mean that all those reviewers are making huge mistakes 😉

[daniellei4510]

But you just said algorhythms are checking the images. So which is it? Humans or computers?

Anyway, point is, you uploaded a couple of images for review and we told you what was wrong with them. You disagreed. That's your perogative. Your images were rejected and there is nothing you can say here that will change that. End of discussion, at least from me.

[Svetlana28639737zwpi]

@daniellei4510English is not your native language? Mine neither but i do understand it.

You didn't quite understand what i meant in regards to MP studio.

Anyway. As i said already. Every algorithm is flawed and adobe has no problem in accepting "questionable" images. Why would it be a problem when they sell.

Other "things" are in question here.

But more distrubing is that you called out one of the top AI accounts for giving AI a bad name.

Makes me wonder about that star in your profile pic 😉

[Jill_C]

The fact that you can find flawed AI images in the database of accepted assets doesn't mean that Adobe should continue to accept flawed assets. In fact, Adobe has expended a considerable amount of effort over the second half of 2023 in trying to clean up the database and eradicate many flawed images. The simple fact is that Moderstors have become better trained, more experienced and much more adept at spotting AI errors. What might have been accepted months ago is no longer clearing the quality hurdle.

[Svetlana28639737zwpi]

Are you reading the posts? Go to the adobe stock and look at the latest additons.

Absolutely every AI image has flaws that you wrote here about.

If they didn't. Well that would mean a global shift in some industries.

It will happen eventualy, perfect ai images in high resolutions. But not at this point.

[daniellei4510]

Here we go again. Been here. done that. A new user posting AI wants reasons why their AI assets were rejected, we give them, and then they argue with us. Yes, a lot of crap AI gets accepted. It shouldn't be. And once a buyer complains, those images will be removed and, quite possibly, the contributor willl have their accounts terminted if there are too many complaints.

Your images were rejected. Fix them if you can and resubmit. If you can't fix them, trash them.

[Svetlana28639737zwpi]

New user? 🙂 Like i wrote to another poster. I am in microstock since 2009. And not with cats and sunflowers 3:)

This sentence tells me about the quality of your experience with stock sites "And once a buyer complains, those images will be removed and, quite possibly, the contributor willl have their accounts terminted if there are too many complaints."

We keep a separate chanel for some images (with questionable qualities) from the stocks sites that have almost become memes in stock comunity.

Not one was removed. It simply doesn't work like that for the numerous reasons.

[Jill_C]

Yes, there are other flaws, but I'm not going to take the time to point out each and every one. The flawed door is easily visible at 100%. We assume that Moderators regularly  zoom in between 100-200%. 

[Svetlana28639737zwpi]

I try to be polite when facing posters like you.

But did you realy think that i have taken any of your "points" seriously?

After looking at accepted AI images daily? 🙂

By your "standards" non of the ai images would and up in the adobe library.

Like i wrote to the other poster. "Just sort the latest ai images by "most recent" and nitpick them like my images here. 90% of them would never got acepted."

[Jill_C]

I have in no way been impolite to you, and if you have construed any of my comments as impolite it was certainly not my intent. I come here every day and invest a considerable amount of my time in trying to help other Contributors, and I always treat them with kindness and try to provide honest, constructive criticism based on my experience of looking at many thousands of all types of stock assets.
1. You came here for advice on rejected images.

2. Several of your fellow Contributors have provided their opinions.

3. You have rejected those opinions in an argumentative manner.

4. You can choose to accept or reject such input; but nevertheless, the Moderation team, who review thousands of images every day and have far more experience in doing so than any of us, have rejected them. None of your arguments will alter that fact.