Should You Be Cautious of Machine Learning and AI for Culling and Editing?

Should You Be Cautious of Machine Learning and AI for Culling and Editing?

Whether you are a professional or amateur photographer finding an efficient way to process your photos is essential. AI can help us do just that, saving us hours of the most tedious part of photography.However, there are reasons for caution.

Time is something we never get back, and carrying out mundane tasks makes us realize the importance of getting on with the important stuff in life. As I make my living from photography, time is also money.

I'm already really busy. Therefore, the opportunity cost of time-consuming culling and editing eats into my bottom line and my life. I've turned away work because I have been too busy processing what I've already shot. I recently canceled going out to shoot the glorious golden light all because I had too much work on my plate. Worse still is having to shave off time with family and friends because of my workload.

In the latter half of last year, I had a particularly heavy deluge of work. Weddings and other photoshoots threatened to overfill my already bulging working days; I often work from 5:30 am until 6:30 pm or later. That is a situation many self-employed photographers find themselves in. So, new ways to streamline my workflow became essential.

With AI, jobs that used to take days now take just a few hours, freeing me up to do other things. Let's look at how it works for me and see if it can work for you.

The Two Problems That Need Solving

The first challenge of any professional photographer is culling photos. When I shoot a wedding or other day-long event, the photos can end up in the thousands; far too many.

Secondly, since no two pictures are the same, I’ve found Lightroom presets never quite give the effect I want on every photo; each photo needs individual development. They can be a good starting point but, after applying them, I invariably must adjust and re-save the preset, and I end up creating ever more of them. I might need to tweak those yet again for the next photoshoot.

Putting Your Trust in AI

As a photographer who is very particular about the products I deliver, it was a daunting thought to hand over some of my tasks to a suite of AI tools. I am not usually one for passing on my creative tasks to even trusted third parties and giving it to a machine was even more scary. Asking an AI system to do some of the heavy lifting grated against the Luddite in me.

However, once I let go of that fear, I found with AI programs I still had absolute control over my photos. I now think of it as akin to having another me looking through my images, helping me decide which photos to choose and what adjustments to apply. My trust in it has grown.

Imagen's is just one of the apps with powerful AI-drivemn tools.

How Do AI Programs Work?

Culling

I haven't yet met any photographer who enjoys the culling process, and I certainly don't. Working my way through 2000+ wedding images takes a long time, and it can become monotonous and easy to make mistakes. I had previously poured over sets of similar-looking shots, comparing them, and trying to find the best ones to keep. But sometimes, I still missed those drooping eyelids, photos with slightly missed focus, and movement-blurred hands on my first and second reviews.

AI-driven programs sort through the pictures saving time and the boredom of doing so. Good programs should quickly identify issues like missed focus or closed eyes, but they should also learn to mirror your selections, choosing the most pleasing shots and rejecting those that don’t come up to your mark.

At first, I would not allow the AI to do that, relying solely on my full control over the program's choices. But placing trust in the programs is essential for that time-saving. Furthermore, quality programs are enormously accurate, speedily weeding out the images that you would want to reject if you were doing it manually.

After Culling Comes Developing and Editing Your Images

When you have culled your photos, you want to apply your developments. Good AI programs have machine learning. They use your previous edits to create a profile that matches your style. That often requires you to identify 3,000+ images that have been developed by you previously. Some programs will learn from the developments you do in Lightroom Classic. 

Once the AI has learned all the subtle nuances that you alone apply to your images, the apps then analyze your new photos and apply individual adjustments for each separate photograph.

Using Other Established Photographers' Profiles

But what if you don’t have 3,000 edited photos to upload that a program can learn from? Luckily, most programs have built-in profiles from other photographers that you can call upon. You may also be able to teach it using a preset that reflects your preferred style. Some programs will also ask you to answer a handful of questions about your editing style. Following that information, the program can create a profile for you.

Applying a Lite AI profile.
AI programs also have the option to fine-tune adjustments too, so you can amend the results if the program doesn't quite give the outcomes you want.

Additionally, every time you add further developed photos to the program, you improve the AI’s understanding of what you are trying to achieve. In other words, the more you use it, the better it gets.

Enlisting the Help of Established Photographers

Some programs will supply other established photographers' profiles that allow you to apply their style to your photos. This is useful for those who struggle to achieve a particular look. You can also tweak these third-party profiles to produce a style closer to yours.

Good AI programs will feature photographers' profiles from various corners of the world, thus encompassing different regional styles and the multitude of skin colors that make up humanity. I can see this feature being incredibly useful for new photographers who are still finding their feet.

Some of the talent profiles offered by Imagen.
 

The streamlined AI development process saves me time. Hardly any manual tweaking is necessary. The programs accurately make adjustments so each of my photos matches my style, and I get a consistent look to the entire photoshoot.

A Caveat

AI is an incredibly useful tool in our arsenal. Whether it is built into the camera to help focus on particular subjects or part of a software package to speed up our processing, it can help us to succeed. I know a novice photographer who uses their camera's AI Subject detection and they achieve multiple successful shots of fast-flying birds that would have been difficult for a seasoned wildlife photographer to capture but a few years ago.

However, I would urge a degree of caution. Although I applaud any tool that makes photography accessible to more people, if we hand over all our creativity to AI, we lose the skills that were once hard-earned. I would urge new photographers to use AI as a gateway to access our art, and also to hone their photography and processing skills skills.

Of course, we can argue that has happened in the past with the advent of TTL metering, semi-automatic metering modes, autofocus, and single-click image adjustments and filters. However, the loss of skills is something to be mourned. Having full control over your work brings a great deal of personal satisfaction that is lost when we hand it all over to automation.

So, although I happily use AI to speed up my workflow for commercial shoots, when it comes to personal projects AI is excluded from processing. In fact, I am regressing to an entirely analog and manual workflow, which I find far more rewarding.

AI Programs With Machine Learning

I think that programs that use machine learning are superior to those that only use pre-programmed AI. There are increasing numbers of AI programs with machine learning out there. Although I used Imagen for this article, Aftershoot is another that I found works excellently for culling and developing. ON1 Photo Raw has machine learning for developing, image enlargements, and other functions.

Do you use AI in your workflow? It would be fascinating to hear your experiences in the comments.

Ivor Rackham's picture

Earning a living as a photographer, website developer, and writer and Based in the North East of England, much of Ivor's work is training others; helping people become better photographers. He has a special interest in supporting people with their mental well-being through photography. In 2023 he became a brand ambassador for the OM System

Log in or register to post comments
11 Comments

AI in the commonly used sense is based on deep learning which is in fact a class of machine learning. Most of the internet sources confuse the termini.

I am continuously trying to figure out how best to ensure I use AI most effectively. In my case [not everyone's, but definitely mine], I fully endorse machine language AI that does not create new pixels while allowing automation of what would be my traditional manual processes; at the same time, I personally am opposed to generative AI that creates new pixels/data, and I want to be sure I do not inadvertently include any in images I create.

This is not a position where one is good while the other is bad; it is my choice for my photography.

My concern is how I keep my images free from generative AI—even in my post-processing. Most of my internet searches turn up many generative AI responses, none of which apply to my question.

I want to be able to say that a given image is mine and is a product of only my actions.

Interesting read thank you.

Cheers, Tessa.

There will come a time when guests at any event will be busily taking a record of everything that is happening. Software will then combine all the images and videos taken by the guests and AI will rearrange every shot taken into a professional looking online record where everyone can view the results as long as you hand over your payment details, you will of course sign over your rights to any and all images taken so this can be achieved. When you take out the human feelings out of everything you end up with a cold unfeeling world, fake smiles and all.

I was trying out Evoto. I'm impressed with it enough I'll be doing the subscription after I've used up the trial credits.

I really like the retouching aspect of it such as reducing/removing acne, blemishes, dark circles, neck wrinkles, smile lines, etc. That's something I would have to do on literally all the images.

I also like how consistent syncing the adjustments (retouch, color grading, exposure, white balance) are from image to image(s).

It doesn't do any cloud uploading/learning. It's just smart programming.

Lastly, it's pretty fast. And in terms of AI subject/background detection, way faster than Capture One and just as fast, if not slightly faster, than Lightroom.

Edit: The downside is the credits expire after 1 year, which they don't disclose until you are ready to buy. Yeah, they'll be getting an earful from me. :) Their lowest plan is 1200 credits for $83.99. I only edit about 350 images a year. Luckily, for the time being at least, there are promotion plans out there that are 500 credits for $36.99. It's unfortunate they don't offer this upfront on their site. With all that said, I'll be sticking with them for a bit because their software is useful for my use.

Thanks. I'd not come across Evoto. I'll give it a go.

Having an AI to do the culling for me would be extremely helpful, but I just know I'm unable to rely onto that fully and I'd still want to do the culling myself just to comparу the results, haha! However, I really like AI-powered tools programs like Photodiva use for portrait editing and I'm looking forward to trying programs that actually create whole editing styles. One of my friends who's also a photographer told me once that the actual photoshoot is the easiest and shortest part of the work and he is absolutely right. Having something to make the most boring and repetitive parts easier would be amazing.

I agree with your friend's assertion. Thanks for the comment

Interesting. We use Aftershoot. It works very well.

The handwringing by photographers over AI is still yet to run its course.
But AI is coming and eventually will be as good as humans in the case of image delivery.
And I would ask 3 questions
1) Why do photographers see themselves as "artists' creating images rather than craftspeople or engineers who deliver the best images using the most efficient tools?
2) Why don't photographers focus on what is "good enough quality for the client" using AI tools rather trying to deliver "good enough for my own standards".
3) What else can AI tools do better and more efficiently than photographers?