"quality issues" - Adobe's screening AI should be trained in chemistry!

[MedBioChem]

Hi Adobe and everyone, 

Recent rejections of my photo and video uploads for "quality issues" seem to be due to Adobe's new algorithms for detecting "blurry" or fuzzy photos. 

While the vast majority of low-quality photos fall into this realm, my photos of molecules (mostrly drug and drug target proteins) are represented by sticks/ ribbon reps surrounded by a transparent surface. This is the standard (and beautiful!) way to present molecular structure connectivity (sticks/  ribbons) and volume (surface) in chemistry. The transparent surface may have led the screening AI to think it is blurred. (see attached, first 3 photos of a human cancer protein).  

Another way to represent how proteins move is to show different strictures together in one photo, basically like a time-lapse photo. Rightmost is a platypus venom protein moving in water. AI must have thought it to be blurry as well. 

Adobe's AI is clearly not versed in chemistry lol. Is there a way to get my uploads re-screened?  

Maybe Adobe should hire some chemists too. 

I already have had 800+ molecule photos and videos approved in the past. And all my photos and videos are of NATIVE 4K resolution (4096x2160) and none are pixelated or upscaled. 

Example: HIV-1 protease 3D protein molecule spinning 4K
https://stock.adobe.com/ph/stock-photo/id/577874247

Thank you.

[Test Screen Name]

1. All images are screened by real humans. That's why it's taking so long.

2. It isn't their job to be specialists, and they never will be. Pictures are judged on their own merits in all cases. The recent flood of poor quality AI generated images means that the vigilance for blurry images has probably increased.

3. There is no appeal. If you resubmit you run the risk of getting blocked.

There is no way they will accept this industry standard blurriness. This is not an organic chemistry library.

[George_F · Accepted solution]

This appears to be a low resolution screenshot of rejected assets, and I'm not sure we'll be able to give any meaningful advice about them.  Feel free to upload the same sized assets you sent to Adobe if you'd like a fresh perspective on them.

Photos are reviewed by real people at Adobe, and not AI.

I suppose an illustration depicting something very technical can be both scientifically correct and also a technical flaw based on Adobes criteria.  In these cases, Adobe tends to prioritize visual technical attributes of photos first.  Rare plants or animals for instance, would still likely be reiected if the photos are of poor quality based on their standards.

I hope you have better luck with future submissions 🙂

[MedBioChem]

Thanks, very helpful. Yes, the screenshot was of  lowres thumbnails.

There is always a market for molecule images regardless of what some think.

Here's one that just sold today, at native 4K resolution (indicated in the screenshot as 4096x2160). The surface is transluscent and shows the molecular connectivity inside.

Fluvastatin, statin drug against high cholesterol, 3D protein molecule 4K

[MedBioChem]

And here's a closeup to show the sharpness at 100% magnification of the top corner part of the above asset - it is teachnically good.  I have sold quite a few of these types of images and videos.

[George_F]

I have no doubt that your accepted assets are of acceptable quality.  Feel free to post your rejected assets as you sent them to Adobe if you'd like us to evaluate them 🙂

[MedBioChem]

Here's the rejected human cancer protein at 800 px and at 100x magnification of the left part.

And thanks - all my previous assets have been approved and have sold - and they are all similar to these rejected submissions. Thanks for the vote of confidence.

 

[MedBioChem]

And of course, I submitted it ^^ at full 4K (4096x2061) resolution. 🙂
I do think the transparent surface was misinterpreted as blurry.

[George_F]

I won't be able to give you anything meaningful without the full sized file that you sent to Adobe.

To confirm, the appearance of being blurry could be enough for a rejection.  Or there could be something else going on that you are perhaps missing.  Even the best of us miss things sometimes 🙂

[MedBioChem]

Thanks. Could be any or all of the above. 🙂

I just wish their screeners can be made aware that molecules need to be displayed this way and not flag them as blurry.

I guess Adobe reads the forums?

If so, please make an exception for transparent/ translucent molecule photos and videos. Your buying clients will appreciate that. 🙂

Thanks!

[Hope springs eternal ...]

[Abambo]

I just wish their screeners can be made aware that molecules need to be displayed this way and not flag them as blurry.

By @MedBioChem

What makes you think that this is the problem? Are they blurry?

[MedBioChem]

Hi, I have submitted (and sold) similar photos in the past. 

They are not blurry per se, more like transluscent, but have apparently been tagged as blurry.

[Abambo]

Hi, I have submitted (and sold) similar photos in the past. 

They are not blurry per se, more like transluscent, but have apparently been tagged as blurry.

By @MedBioChem

Tagged by whom?

[Abambo]

Here's the rejected human cancer protein at 800 px and at 100x magnification of the left part.

By @MedBioChem

What we need is an asset as submitted (100%, even if you use photoshop to crop).

Look at aliasing/antialiasing and look at your tube connections. 

And do not assume, the moderators understand, what they see. If something is inconsistent, even if it is according to your understanding, correct, it may earn a refusal.

[MedBioChem]

Thanks - the quality all checks out.

Agree with the last point.

[Nancy OShea · Accepted solution]

Read Me First: Stock Reviewers Decisions are Final.

Adobe Reviewers NEVER read these user-to-user forums.  That's not their job.  Their only job is to evaluate thousands of assets as quickly as possible.

Speaking as Stock customer who also did media preparation for brain researchers, I know from personal experience that excellent print quality is #1.  When selecting graphics for print publication, posters and video presentations, edges must be sharp & well defined.  No way around it.

Your images will not print well.  That's a shame because aside from that, they're pretty good.

Hopefully you can fix the problem and resubmit.

Best of luck on your next submission.

[MedBioChem]

Thanks for the insights. I'm not sure though about them not being printable to an acceptable quality - print journals routinely have images like mine, and do not even require 4K resolution (see attached). And even for large posters, 4K is more than enough resolution. I do appreciate your points, though.

What is the state of the industry regarding print, though? I am under the impression that it has declined significantly now that most clients prefer digital over print. My photos still do sell, and there is a market for them, albeit not as large as say, human model photos. 

As for screeners not reading this forum, that's their prerogative, but change usually starts from a small request, so who knows? Best to ask and try and find out for sure. 🙂

[ricky336]

Hello,

Moderators reading this forum is rather irrelevant. When submitting one has to think about how the image can be used - the ability to print the asset is one of many considerations - even though these days it might not be so popular, people still print.

And another point is that when it comes to colours, do the printed colours match to what is shown on the screen? I can imagine when representing molecules, colour is very important.

[MedBioChem]

Thanks, yes that is a consideration, and an important one. Color (are you British? "colour" 🙂 ) is important but for molecules, what's most important is the difference between atom colors. Color space/ calibration is no longer an issue with most web/print applications. [I do remember the early days of the web when we would all be designing for 256 "web-safe" color spaces lol). Any decent graphic designer can always change the colors if he/she so wants to, so I don't think this is the issue.

Anyway, this is just a small subset of my submissions, but I do love clarity.

[Abambo]

As for screeners not reading this forum, that's their prerogative, but change usually starts from a small request, so who knows? Best to ask and try and find out for sure. 🙂

By @MedBioChem

That won't happen. And there won't be a recourse method in the near future. Every rejected picture would land there, as everyone is convinced that the "Adobe AI made an error". As you don't want to post the full res picture, which is your good right, you will need to check yourself the asset. Just simply don't assume you know why the moderator rejected the asset. For sure, it's not the transparency turning blurry. That does not happen.

I've worked long enough in 3D, to know that sometimes elements do no behave as expected, surfaces turn up unexpectedly, a light goes out unexpectedly (for animations) ... Check your asset, if you do not find an error, check the histogram. That could be a good indicator. If nothing turns out, you are either lost in your assumptions, or the moderator simply had a bad day. Resubmit after a while, best with some modification. A different moderator may have a different eye on your assets.

[MedBioChem]

Good points!

All photos and videos I have submitted were rendered in software with ray-tracing and are consistent per frame for all frames. Histogram checking shows they are well-behaved. No clipping either. Nada. 

I guess we differ in opinion. FYI, similar assets have sold in the past here at Adobe stock, and I have used the same software, settings, and procedures for all of them, including the recently-rejected ones. Maybe it is a new reaction from Adobe screeners to the wave of poor images from AI programs. Or a bad day. But who really knows?

Thanks for your tech-savvy comments.

[MedBioChem]

"For sure, it's not the transparency turning blurry. That does not happen."

To be clearer, the transparent surface *appears* blurry - it doesn't *turn blurry.

"For sure" ... "That does not happen." I don't really know if it does or does not and I need further convincing otherwise.  🙂

[Abambo]

To be clearer, the transparent surface *appears* blurry - it doesn't *turn blurry.

By @MedBioChem

It even does not apear blurry. I've done 3D. I never had these appearances. Render something, post it here. It can be a crop of a render, but it needs to be 1:1. 

We only believe what we see. 😏