Main objections for automation of photos culling

Introduction

Since no more than couple of years, a few software capable (or claiming being capable) of culling automatically images thanks to AI and different algorithms or models have been released or are getting close for being commercially available. As usual, some will be a better match to your needs than others. But it is interesting to note how doubtful many people can be when it comes to the principle of automating photo culling.

The point of this post is not to compare the software – I cannot do it as the owner of one of this companies – but to give a quick overview of the main reasons for people to reject such tools. What matters for now is not that the arguments are valid or not, it is just what many people believe so far.

And let us be honest, all these points must be addressed properly to convince photographers to use such software massively.

A compilation of valid points to be addressed...

…Either because there are truly valid, or because the perception
there are valid is certainly strong.

  • “Composition, aesthetic quality and or message is in the eye of the beholder, NOT the AI of the computer.”

  • “Solution in search of a problem”

  • “For photographers who hate the art of photography”

  • “if you are a boring photographer this will work if you are a cutting-edge photographer the app will remove your best photos!”

  • “Are you sure you want someone else anonymously and without critique to form your view on the world? I get it if it is restricted to purely technical/form elements like sharpness vs. overall blurry mess, color harmony, some compositional elements like distracting highlights, tangential lines that obscure subjects, noisy backgrounds etc. Even then some photos one would consider great may be rejected by the app…..”

  • “Do they have an app that views your images so you don’t ever have to look at them? 🙂 I think the problem with this is that there is no way an app could evaluate sentimental or emotional value of an image to an individual.”
 
  • “As far as I can think, AI culling could be useful for very few scenarios, the only one coming to mind being high volume, fast turnaround studios or pop-up studios that shoot people.”

  • “Spare us from this false belief in artificial intelligence for evaluating our images, it’s just pushing the world into a look of sameness, where every person in every photo is happy and smiling. Stepford wives. Never mind that there might be a reason for closed eyes — meditation, grief, deep thought,… . And probably no consideration of what might be off to the side — perhaps the last photo of your grandmother.”

  • “The future of images? Having a algorithm choose what photos meets it’s criteria. Welcome to the new world of plastic same-o same-o images. Sort of like going to a workshop and having everyone stand in the same spot to shoot the same image. Blandness prevails, all hail blandness.”

  • “Wow, I can imagine future venues with several dozen cameras taking random shots and then preselecting the best ones. The question is, do we really want to go down this path of giving everything we do to technology? There must be a happy medium somewhere and I wonder if in some areas we have already crossed it.”

  • “When I have a set of images and want to choose the best/sharpest one, I just put them into Lightroom compare mode in pairs and delete the worst of the pair. Repeating this on successive culls is a pretty fast way to narrow the choice down to just 3 or 4 shots for final editing.”
  • “How is software going to know which are the most important attributes to a photographer or publisher? Would it e.g. prioritise sharpness over composition? If you were shooting a fast moving sport in which the positions of the players constantly changed, how would it know whether you had the right player and the ball in the position you want? I can only see value in a product like this if it is fed 1000 almost identical images, in which you have already satisfied yourself that the composition, lighting etc are as you want, and just want to eliminate images with certain attributes. My point is that AI won’t know what *you* want.”

  • “The idea that any photographer would turn over any decision making to AI software seems incredibly ridiculous… let alone paying a monthly fee to do it. Can you have it just click the auto adjustments too?”

  • “No! Technically good photos are not always the best photos. Why would I want a machine to think for me?”

  • “It’s not, because you still have to look and see what was the best image. The best image it’s not always the sharpest, most contrasty and with lower noise.”

  • “It’s nice to have but I’m not sure if I should trust machine to do this kind of stuffs. Slightly oof photos, off exposure photos but with good composition and expression are better than perfect technical photos but with bad composition and expression.”
 
  • “Much of my sports pictures are culled based on 1st focus, 2 interest, 3 action. BUT, interest and action could change based upon how it’s cropped. Would the AI know that? Would it know the difference between action sports and athlete portraits?”

  • “Let someone (or something) else choose your ‘best’ photographs” sounds, to me, too much like “let someone (or something) else just take them in the first place.”. I’m sure other people see it differently, which is fine. But personally, I got into photography to actually *do* it. The whole thing. To have ideas, to plan, to shoot, to cull, to edit, to deliver, to look at what I did. Again: no judgments, but personally I have a hard time being interested in any technologies that push *me* out of *my* art. Also, I’m sceptical that someone just blitzing through a take in Photomechanic or even a Capture One session wouldn’t end up getting a faster final cull than the person who imports to Lightroom, generates previews, engages this system, and then reviews-approves-or-denies its various selections. Perhaps it will prove me wrong?”

  • “As a hobbyist, I don’t feel the need for auto-culling. Cleanup is part of the process where I learn how to be better and to have a critical eye for the content, but it’s also where I can learn what can be saved in post and what its a lost cause. I’ve also heard from others that “sharp” isn’t the only acceptable form of a good photograph, so it would be sad for AI to trash what might be a nice moment because it wasn’t sharp enough.”

  • “Nothing here convinces me there’s a shortcut. And there are trust issues. Reminds me of OCR and AI that would summarize text. They can be 90-95% accurate, but what gets lost in the 5% can be killer…think dropping a zero in a contract. Ditto for photos; something unique might be culled when it’s exactly the kind of risky shot that actually worked. We’ve all probably blundered into those. And in any case, most culling is like speed reading. I can chuck the obvious lens-cap-on and badly out of focus shots so quickly it’s not work trying to automate. Similars are hard, but they will be anyway, since slightly different framing, exposures, DOF, etc was used on purpose. Not to mention HDR, focus stacking shots, panos.
    But I have found Fast Raw Viewer to be really helpful in culling and ingesting. Not AI or automatic, but well designed just for the task of culling raw images.”

  • “The thought of auto culling throws a shiver down my spine. My main fear is that a shot might be out of focus but that might be intentional, a shot might look wonky but I might have done it for a reason. Auto culling is too risky for such things. Generally full automation in any art form is not a good idea. Art by definition is a thing of perspective of individuals, there’s no right or wrong. So the appreciation parameters are different for different people. “
 
  • “Personally, I’d never trust a computer to determine if an image is acceptable or not. For years, I’ve used a three pass method where each pass takes longer per image, but has fewer images remaining. In the old days, I used the same basic process on prints from film, but here I’ll describe how I do it in Lightroom. “Before starting, I recommend at least 24-48 hours before looking at your images. This gives your mind a little time to forget its expectations so you can look at the images for what they are rather than what you expected them to be. First is the pick/reject pass. This is extremely fast because all you’re doing is giving your gut reaction. Only 1-2 seconds per image. The question I ask is “Is this image worth my time?” You’ll eliminate anything with major technical issues, blinking subjects, and anything you just don’t like. On a 1000 shot shoot, I generally eliminate 500-600 of the images and it takes only 15-20 minutes. Next, filter to only show the pics, select them all and mark them 3 stars. The work through the group and ask yourself “Is this image better than my average or worse than my average?” I keep my fingers on 2 and 4 and either leave them alone or increase or decrease the rating accordingly. This process takes about 10 minutes and I end up with about 100 3 star and 200-250 each for 2 and 4 stars. Finally, filter for 4’s. These should be the best shots you got, and now you look through them and ask “Is there anything special that sets this image apart?” I’ll generally revise a few of them down because they aren’t up to scratch, but generally about a third of them will be marked 5 stars. When all is said and done, I preset the 5 star images to the customer along with anything in the 4’s that I need to get proper coverage. In a pinch, I might grab a 3 if it’s something too important to miss, but in the end I might present 75-100 images to the customer per 1000 shots and the culling process takes less than an hour. “

  • “The vast majority of Canon camera owners DO NOT consider themselves artists. If you are an artist, then this product is NOT FOR YOU! Simple as that.”

Sources

1 thought on “Main objections for automation of photos culling”

  1. Pingback: Automation of photo culling: why detecting Eyes closed is not enough, by far – Tristan Romain Renaud

Comments are closed.