I would assume that its about the unsharp Details in those pictures. De Chip does not have much Details and also the Bowtie have unsharp areas.
Before you submit, please review the submission guidelines carefully and compare your work with other Stock inventory. To be accepted, your work should be as good or better than what's already represented in your keyword category.
Are these AI generated Illustrations? They appear to be. I wonder what the reason is for upscaling through Gigapixel?
At first glace the cutout work on the bow tie needs some work to smooth it out a bit in places, and both the computer chip and motherboard images have an odd overprocessed appearance in the details when zoomed in. This is very common with AI generated content, but it could also be from Gigapxel.
I doubt its the cutout work though. I do that part manually, not a background remover, which I find dont do a great job. I might have been under the impression that Gigapixel AI improved quality, rather than just enlarging the photo...still not sure.
I wonder if generating images in certain styles would create better assets. For example, cartoon drawings as the colors will be flat throughout. Same with line drawings, etc. Basically areas where the ai doesn't fall short of actual photos/illustrations.
I easily see imperfections in the bow ties. I didn't look that closely at the computer images, but I suspect similar problems. I think I'm still at only about 80% accepted AI images - still figuring out where the quality standards are.
Upscaling is not allowed. I'm not sure what gigapixel is/does, but I use an AI "image restoration" to make larger images (SwinIR). I've had some AI that imitates traditional art media, some that imitates digital art media, and some that imitates photographs accepted, but it has to be 100% correct. In fact, I've been a traditional media artist for decades, and I'm starting to think it's just easier to make traditional art than acceptable AI art. (I'm having a great time playing with this new technology, though.)
AI images can't have anything that doesn't "make sense" - that a photograph wouldn't capture or an artist wouldn't include.
It was just the standard reason/no-reason of quality issues.
So I don't know enough (outside of the guidelines given on the sites supporting documents) about the factors of quality to know which would be the issue.