Well, that is astonishing but on the other hand I watched a training video from SS years ago and they said to make a vertical version for the magazines and a horizontal one for the websites. That's why I was very astonished that the vertical version of my corn image was rejected and it's not even identical to the horizontal version. I am still speechless and really angry.

Years ago is the answer, they changed with the new "stricter" (ha ha) standards. And the fact that some sort of software is detecting the similar. I don't know how some sets get past, but from reading here, the reviews have clamped down on similar elements in images, to the point that fairly different images, can get rejected as similar, because of a identical prop in the images, or same model.

Example of what SS sees as similar images, in a search:



Anyone want to guess what the search was? Purse, Wallet, Box Camera, or one of the various others?

Something grey

Yeah, and the point is, however they determine similar, it's by way of some strange analytical software and not something that you are doing wrong. Looking at the images I thing someone could start to guess it's the metallic dot, the general black shape and all are isolated images.

Then we ask, how does this happen, and pass? I think the guy with the obvious similar images, has this for everything he has up, and over 276,000 images listed. I'll guess he just waits a month and uploads another and the software misses it. Looking at his portfolio by age, there are no duplicates, but click on any image and he has many of everything.




Whatever it is, he's got raw pork knuckle covered?

That's my guess:  How do people manage to avoid the rejection for Identical Submissions?

I did a small experiment today, with a smartphone image of a forest:
-> 3 crops of the same image and a 4th image with only some lightbeams on it
-> all 4 images were accepted in 2 submissions (2 images accepted each submission)

It seems some very similar images just get through...

Whatever it is, he's got raw pork knuckle covered?

Oy way

Quote from: thijsdegraaf on December 16, 2021, 14:22

Once every few months I upload the photo again (mainly out of curiosity if I succeed) but it is always rejected. As if they're thinking: �h... He's trying again!"

make some tiny changes to the photo

Good idea.

[/quote]

Quote from: thijsdegraaf on December 16, 2021, 14:22

I had once been stupid enough to upload two pictures of different gooseberries in different bushes at once. Of course, one photo was rejected for similar. Both photos were approved by Alamy and Adobe. Once every few months I upload the photo again (mainly out of curiosity if I succeed) but it is always rejected. As if they're thinking: �h... He's trying again!"

they do track previous submissions; the instructions say to indicate it's a re-sub to get a different reviewer (not clear whether initial reviewer would see it anyway)

I think so too. I'll give it a try with some changes. Not because this photo sells so well, but I think it would be interesting to see if it works.

Yes to all of that. I was just watching the similar rejections that people posted and sometimes there's one model (distinctive face recognition?) or a distinctive prop or part of a design, and they get rejected for similar, when they really aren't so much.

Who knows how far back the reviewers can see? But that guy I posted just up the thread has 7-10 of many food shots, all of them micro adjustments and then some identical with just the background changed.

After reading, it can be a reviewer, but I'm sticking to my theory that there's some pre-review software and when it misses, then things can get past, and when the Bot sees similar, the image is dead and never gets to a reviewer. Someone just told me he got a 20 minute rejection. Yet most of the time we are seeing accepted images take a day or two.

I'll bet that something with a distinctive subject, and I'll crop portrait and landscape, one of them will get rejected for similar. But the colors and the main point have to be standing out enough for the Bots to recognize them.  Another entertaining challenge would be get a rejection for similar when the shots are completely different subjects, but similar composition and colors.

I mean more than Gooseberries, on two different plants, something like a green striped ball and a gooseberry. Only if I'm in the mood to waste time, for nothing?

Quote from: Uncle Pete on December 17, 2021, 10:04

...
Who knows how far back the reviewers can see? But that guy I posted just up the thread has 7-10 of many food shots, all of them micro adjustments and then some identical with just the background changed.
...

they've re-activated the way-back machine. i recently reworked a number of images from 10 years ago, using current apps to reduce noise,sharpen, etc - most were accepted, but a few were tossed as similar as they'd been accepted long ago, and the ai/reviewer didnt consider there were improvements

So they are now searching through "377,304,034 stock photos, vectors, and illustrations" ? That machine must be working hard to create identifications by major features, colors and shapes. I was always amazed when I used Shazam at how fast it can identify a popular song. Funny also is how many older songs aren't in the database. It must use the key first, to start the elimination... and without going on about that, I can imagine the SS duplicate finder, or similar analysis, does the same, starting with (for example) major color, then shapes, and so on down whatever hierarchy they have devised.

A few years ago, I added some backgrounds, like wood or a solid color, to old isolated images, to make them better. They aren't, they don't sell but they were accepted.

Amazing variety!  Had to scroll down to get it all...

...
So they are now searching through "377,304,034 stock photos, vectors, and illustrations" ? That machine must be working hard to create identifications by major features, colors and shapes
..

rejection for similars is only against the artist's portfolio, so a much smaller set - a couple of quick google reverse image searches took about .5" each, against an enormous dataset (not including time to upload the image); so, checking 300K images/week at that rate would take only about 2 hrs/day. a search for 'similars' on a buyer's search otoh uses metadata

Quote from: Uncle Pete on December 18, 2021, 11:21

...
So they are now searching through "377,304,034 stock photos, vectors, and illustrations" ? That machine must be working hard to create identifications by major features, colors and shapes
..

rejection for similars is only against the artist's portfolio, so a much smaller set - a couple of quick google reverse image searches took about .5" each, against an enormous dataset (not including time to upload the image); so, checking 300K images/week at that rate would take only about 2 hrs/day. a search for 'similars' on a buyer's search otoh uses metadata

Yes, I understand the time factor. I'm still surprised that they are checking back, yours against yours. And I never bought the keywords were a trigger because I know that I have events with the same words, except the drivers name (as one example) and they don't get rejected for similar, but the same car at a similar angle, in a different place, where obvious colors and shapes match will get rejected. The AI is reading colors, open spaces and shapes.

I recently got a photo of two totally different mountains rejected as beeing "too similar" *sigh*
It wasn�t on shutterstock since I closed my account with them in 2020 and I forgot which agency it was- but I submitted lately two pictures of two different mountains in the italian dolomites (and properly put the different names in the titles) and one got rejected for beeing "too similar" to the other....
Come on - sure a mountain is a mountain and they sometimes look alike, but those were two different mountains in two different places    And I don�t shoot them for plainly beeing a nice mountain but to be that specific mountain....

Yup, that's how the machines see similar, shapes, open spaces and colors. I can imagine you could upload a mountain in Italy and another in Scotland, and if they are close to the same peaks and valleys, it would get rejected for similar.

Now about the Pork Knuckle Guy?



How does he get away with this? But you get rejections.

Quote from: Uncle Pete on January 30, 2022, 18:11

Now about the Pork Knuckle Guy?

47 of the same red ripe big juicy tomato isolated!

Yes, but other people including Cascoly get rejections for "similar" years later, with a variation of an image. I looked at the collection of Mr Pork Knuckle and his are not all from one upload, the file numbers, or if anyone cares to look, are spread out overs years. I'm guessing he does a session, then uploads one. Does another session, uploads one... Etc. but saves the series and then dribbles them in, one at a time at a later date. He has many repeating images, that are uploaded over a long period of time. I didn't see the descriptions being different either.

Yes, some day I'm going to make the perfect sliced tomato image and take over the top of page one! I'm still working on the most perfect cheeseburger and fries (they are a better combination?) and some days I add fried eggs, Plop and Shoot. 

910,053 sliced tomato
120,447 cheeseburger and fries
555,458 fried eggs

These are things that anyone serious should actually avoid and not waste their time, but people keep shooting and shooting and wonder why their RPI isn't so good?

And seriously buyers don't do a search like that, if they have any sense at all. More than one or two words is a more realistic way to find your placement.

Example:
1,474 fried eggs bacon muffin (page 3)
"fried eggs bacon muffin overhead"  (page 1)

50,171 big chicken - both of mine on page 1.  https://www.shutterstock.com/search/big+chicken

...

Yup, that's how the machines see similar, shapes, open spaces and colors. I can imagine you could upload a mountain in Italy and another in Scotland, and if they are close to the same peaks and valleys, it would get rejected for similar.

Now about the Pork Knuckle Guy?...

on the shape & color theory - most of those have large areas of different colors - otoh, i had similars rejects for a series of whimbrels on the coast - obviously different - # of birds, positions, etc - BUT background of surf & sand similar

afa 2 mtns, i've never seen that - perhaps there's2 steps -1st find similar colors, etc, then check if keywords are same? (Could still get rejected if they have 50 tags mostly duplicated