AI QC?

[JamesOS]

Hi,

Wondering if Alamy has started using some form of AI to QC images?  I've just had a perfectly good image rejected for a spurious reason: odd.

[Alamy]

On 17/11/2023 at 14:00, JamesOS said:

Hi,

Wondering if Alamy has started using some form of AI to QC images?  I've just had a perfectly good image rejected for a spurious reason: odd.

Hi James,

 

No we have not started using AI to QC images, and we currently have no plans to replace our human QC Team with AI! 

 

If you want some further clarification on why a submission failed, drop us an email at contributors@alamy.com and we will be happy to discuss. 

 

Thanks,

Louise 

[JamesOS]

On 20/11/2023 at 13:23, Alamy said:

Hi James,

 

No we have not started using AI to QC images, and we currently have no plans to replace our human QC Team with AI! 

 

If you want some further clarification on why a submission failed, drop us an email at contributors@alamy.com and we will be happy to discuss. 

 

Thanks,

Louise 

That's good news thanks Louise!

 

I did query why you'd failed one of my images but nobody replied; so I resorted to this forum.

[Steve F]

8 hours ago, JamesOS said:

That's good news thanks Louise!

 

I did query why you'd failed one of my images but nobody replied; so I resorted to this forum.

 

Hi James, you could always try posting the image on the Forum at 100%, together with the QC failure reason.

[wiskerke]

1 hour ago, Steve F said:

 

Hi James, you could always try posting the image on the Forum at 100%, together with the QC failure reason.

Maybe explain:

Post the image somewhere on the internet at 100% exactly like you send it to QC. If you're worried about it being stolen, maybe put a watermark on it.

Then link to it here on the forum.

(The images on this forum have to be publicly available somewhere and cannot be more than 750px high.)

 

wim

[Sandi]

I submitted 10 images for QC. One has failed for a dubious reason. The left-hand column of my Images page now tells me 9 have passed QC, one has failed. This is new - they would have all been chucked out previously.

However, I cannot tag the 9 good images and the display shows each one still to be in QC.

Is this a software update in progress? Or do I need to resubmit the 9 good images, as before?

[Steve F]

2 hours ago, Sandi said:

I submitted 10 images for QC. One has failed for a dubious reason. The left-hand column of my Images page now tells me 9 have passed QC, one has failed. This is new - they would have all been chucked out previously.

However, I cannot tag the 9 good images and the display shows each one still to be in QC.

Is this a software update in progress? Or do I need to resubmit the 9 good images, as before?

 

You probably need to resubmit all, but never heard of images from a failed submission being passed. Don't assume the other 9 are ok, Alamy fails a submission as soon as they find one image that doesn't pass.

[Mr Standfast]

2 hours ago, Sandi said:

I submitted 10 images for QC. One has failed for a dubious reason. The left-hand column of my Images page now tells me 9 have passed QC, one has failed. This is new - they would have all been chucked out previously.

However, I cannot tag the 9 good images and the display shows each one still to be in QC.

Is this a software update in progress? Or do I need to resubmit the 9 good images, as before?

 

Nice pictures Sandi, yes I looked we all do! Bin the the bad image and triple check the remainder. 

Alamy have a useful guides here

https://www.alamy.com/blog/common-qc-failure-reasons-and-how-to-avoid-them

and here

https://www.alamy.com/contributors/alamy-how-to-pass-qc.pdf

Apologies if you have already seen them, but it's usefull to post them for the next visitor who looks at this topic.

 

As Steve F as said "never heard of images from a failed submission being passed" but thats probably not relevent it's the one that failed that counts.

 

Good luck.

 

[Mr Standfast]

Always amused by two assumptions.

 

1. If PALAMY had an AI would they waste it on QC.

 

and

 

2. The AI is incompetent. 

Edited November 22 by Mr Standfast

[MDM]

10 hours ago, Mr Standfast said:

Always amused by two assumptions.

 

1. If PALAMY had an AI would they waste it on QC.

 

and

 

2. The AI is incompetent. 

 

Or Louise is actually a highly competent AI and knows just what to say to convince everyone that there is no AI on Alamy QC. However, assuming that is not the case and Louise is actually a human employee of Alamy, it's great that Alamy have finally come out and said that there is no AI involved in the QC process.

 

I did wonder at times if there was an automated initial screening when experienced contributors were reporting QC failures that to them were not easily explained. However, in almost all cases when these contributors posted the images that failed at full size on an independent site, it became clear that the failures were correct and Alamy QC were  getting it right. That said, there was a very recent report of a failure where I (and several others) could not see any reason for the the image failing.

[Mr Standfast]

1 hour ago, MDM said:

 

Or Louise is actually a highly competent AI and knows just what to say to convince everyone that there is no AI on Alamy QC. However, assuming that is not the case and Louise is actually a human employee of Alamy, it's great that Alamy have finally come out and said that there is no AI involved in the QC process.

 

I did wonder at times if there was an automated initial screening when experienced contributors were reporting QC failures that to them were not easily explained. However, in almost all cases when these contributors posted the images that failed at full size on an independent site, it became clear that the failures were correct and Alamy QC were  getting it right. That said, there was a very recent report of a failure where I (and several others) could not see any reason for the the image failing.

 

Your first point, scary eh!

 

Second point.  QC are human, mistakes can get through. Making subjective assesments without errors is difficult. My experience  of reliable checking involved constant cross checking, training and assesment, encoouraging taking second opinions and having a robust escalation process. No body gets QC'd more than a QC inspector. Mistakes can get through but we don't see many. A no mistake QC regime would cost more and who pays for that? I'm not griping at you MDM,  just putting down my thoughts for anyone who wants to run down my old career!

 

G'night all.

 

 

Edited November 22 by Mr Standfast

[MDM]

 

1 hour ago, Mr Standfast said:

 just putting down my thoughts for anyone who wants to run down my old career!

 

 

I'm all for QC. I had a few failures in my early days on Alamy and that certainly made me improve the technical side of my photography in terms of image sharpness and making sure there were no dust spots. It was more difficult then to ensure sharpness as images had to be a minimum of 48MB and sharpening was not permitted at all. All my stuff had to be upsized as I was using a 12MP Nikon D700. Only once was I puzzled that an image failed. 

 

My only claim to working in QC was the half day I spent in an onion factory in Holland in my student days a long time ago. It involved picking out bad onions from a conveyor belt of tiny onions destined for pickling. It was vile in every sense and I didn't last the day. Definitely a job for an AI.

Edited November 23 by MDM

[Inchiquin]

But is it possible they might use AI to assist the humans, e.g. by suggesting images that might warrant a closer look (rather like VAR sending the referee to the monitor)?

 

I would consider that an acceptable use of AI which would make life easier but still leave the decision in the hands of the human.

 

Alan

[MDM]

4 hours ago, Inchiquin said:

But is it possible they might use AI to assist the humans, e.g. by suggesting images that might warrant a closer look (rather like VAR sending the referee to the monitor)?

 

I would consider that an acceptable use of AI which would make life easier but still leave the decision in the hands of the human.

 

Alan

 

I did wonder if Alamy QC was using some sort of initial automatic screening process when several experienced contributors were reporting failures around the same period a couple of years ago (can't remember when) but, when people posted their failures, it was clear that QC was getting it right anyway. I imagine it would not be too difficult to program an app that would look for points of sharpness in an image, given that this has been available in Photoshop for some years now (Select Focus Area) which could indicate SoLD images very quickly but we can rule that out given what Louise says above.

[Steve F]

26 minutes ago, MDM said:

 

I did wonder if Alamy QC was using some sort of initial automatic screening process when several experienced contributors were reporting failures around the same period a couple of years ago (can't remember when) but, when people posted their failures, it was clear that QC was getting it right anyway. I imagine it would not be too difficult to program an app that would look for points of sharpness in an image, given that this has been available in Photoshop for some years now (Select Focus Area) which could indicate SoLD images very quickly but we can rule that out given what Louise says above.

 

It is good to know it's not done by a computer programme. There are lots of images that might fail an automatic AI programme, based on Alamy's extensive QC failure list. More 'artistic' images for example might have large areas of clipped black shadow, be high / low key, show motion blur deliberately, etc. I imagine there would be lots of false positives.

[Alamy]

18 hours ago, MDM said:

 

Or Louise is actually a highly competent AI and knows just what to say to convince everyone that there is no AI on Alamy QC. However, assuming that is not the case and Louise is actually a human employee of Alamy, it's great that Alamy have finally come out and said that there is no AI involved in the QC process.

 

I did wonder at times if there was an automated initial screening when experienced contributors were reporting QC failures that to them were not easily explained. However, in almost all cases when these contributors posted the images that failed at full size on an independent site, it became clear that the failures were correct and Alamy QC were  getting it right. That said, there was a very recent report of a failure where I (and several others) could not see any reason for the the image failing.

I'm pretty sure I'm not AI, but I will take the 'highly competent' comment! 😄 We aren't currently looking at using AI for QC, so you'll have to suffer no QC over weekends for the time being, and if you think your images adhere to all the guidelines and have been unfairly failed, just drop us an email and we'll be happy to ask our human friends in QC to elaborate on why it failed. 

 

Thanks,

Louise 🤖😉

[Ognyan Yosifov]

29 minutes ago, Alamy said:

I'm pretty sure I'm not AI, but I will take the 'highly competent' comment! 😄 We aren't currently looking at using AI for QC, so you'll have to suffer no QC over weekends for the time being, and if you think your images adhere to all the guidelines and have been unfairly failed, just drop us an email and we'll be happy to ask our human friends in QC to elaborate on why it failed. 

 

Thanks,

Louise 🤖😉

I'm willing to cover the weekends for QC if at least 51% of my fellow contributors vote for it

with a thumbs up! 🙂

Edited November 23 by Ognyan Yosifov

[Inchiquin]

8 hours ago, MDM said:

 

we can rule that out given what Louise says above.

 

 

Well what she said was "we have not started using AI to QC images, and we currently have no plans to replace our human QC Team with AI".

 

But using AI to QC images is not the same as using AI to suggest possible failures, nor is replacing the human team with AI the same as using AI to help the human team.

 

I'm not saying this is the case, but I don't think what she says rules it out at all.

 

Alan