In this post I’m looking at Artificial Intelligence and trying to answer the question “Does AI improve Photo Processing?” Is it just another smart way of selling software or is it the revolution it aspires to be?
Does AI Improve Photo Processing?
Does AI improve photo processing? Is an AI processed image a photograph? Is AI just another marketing term? Will AI take over the Photo processing industry?
Artificial intelligence is a thing these days. There are Photography processing tools built on AI. So what is it? Does it work? And where does that leave professional photographers?
Let’s take the first of these questions.
What is AI?
The founding fathers of AI, a couple of gents by the name of Minsky and McCarthy defined their invention as “any task performed by a program or a machine that, if a human carried out the same activity, we would say the human had to apply intelligence to accomplish the task.” A fairly broad definition but looking a little closer we probably say that AI involves the following human traits – Reasoning, Learning, Analysis of data and ongoing results, in pursuit of solving a problem.
In terms of image processing, that definition narrows the practitioners of true AI enhanced tooling down to a very few. DxO Photolab 4, built on the datasets compiled by DxO Mark, describing almost every camera body and lens combination imaginable qualifies. Skylum are trailing a new AI edition of Luminar which we will be reviewing in due course. Photoshop is introducing some of the features such as sky replacement that Luminar featured 12 months ago.
Does it Work?
As has been the case with Smart home Technology, AI in Photography has been controversial. The concern with AI in photography is that it will somehow run away with the photograph and return with something that does not represent the photographer’s vision. Taking DxO Photolab 4 as an example, this is emphatically not the case. What DxO have done really well is based on the assumption that the best place to start processing a photograph is the best possible version of the RAW file.
There are three choices in DxO – No processing at all, Optical Corrections only and Basic Processing. The first switches the AI elements off, the second applies the optical corrections that the photographer would probably apply anyway and the third introduces some basic processing. In addition, there are pre-sets that the user can apply to get a certain look and feel.
Optical Corrections in DxO Photolab 4 work brilliantly. It simply corrects the issues that certain lenses introduce – most easily seen in wide angle lenses these features include geometric distortion and vignetting. With the default set to Optical Corrections or Basic Processing the issues will be corrected in line with the characteristics of that lens-body combination. In this aspect we can confidently say that it is true AI and it certainly works.
Luminar 4 boasts some AI in its processing however this works noticeably differently. Luminar is based on processing “recipes” gleaned from hundreds of photographers, so it leans more towards templates in Version 4 than true AI. Where Luminar appears to leverage AI really effectively is in the details that are tricky to get right manually. The obvious example is sky replacement – the lighting on the rest of the photograph has to be adjusted to match the sky. Luminar does this well. It will be interesting to see how far this has progressed in the new Luminar AI version which is coming out soon.
Is a photograph developed with AI still a photograph?
This is a thorny issue if ever there was one. From the perspective of competition entries, some fine art competitions are fussy about Photoshop, never mind whole sky replacements. The normal wording in these competitions refers to such processing as may substantially alter the authenticity of the photograph.
From another perspective, that of a commercial photographer, what matters is the impact of the picture on the audience. Compositing and Augmented Reality (AR) are already finding their way into the commercial photographer’s toolbox so why not AI? I processed the picture at the head of this article using the AI assisted Luminar 4. It was commissioned by a Real Estate agent in Southern Spain and both he and his client were very happy with it.
Another perspective is asked by the established photographer. “Is a photograph developed with AI true to my vision?” I think it’s up to the individual. My landscape photography is principally shot for my own satisfaction, many of the pictures are sold as stock and I know pretty much exactly the look and feel I’m after in any given shot.
When I set the camera up for a landscape shot, I have usually found myself at the location at a certain time in certain lighting conditions as part of a plan. The vision for the photograph is there long before the shutter is pressed. Everything therefore is working towards my vision. However, this happy state of affairs only became possible after I’d spent many years taking photographs often enough to understand how my camera would work in a variety of lighting conditions.
For the photographer who is at the start of their journey, photography can often be “hit and hope”. Get the exposure right, sort out the framing, the optics and the colour balance and you might get a good photograph. For this photographer AI could be an inspiration and a great learning tool. There are times when I simply run out of inspiration and it’s useful to cycle through some presets to see if anything resonates.
What will be the impact of AI on Photography?
Does AI improve photo processing? Many of us have spent years getting to grips with Photo Processing. In my comments about DxO Photolab 4 I’ve described a use of AI that takes time off my processing time. That makes it useful. On the other hand, the picture at the top of the article processed with Luminar 4 took a few minutes to process but I suspect another photographer could get a very similar result. So this poses two questions – “Will all photographs look the same?” and “Will Professional Photographers still have a job?”.
To answer the first question, I cite Instagram where many of the images look exactly the same. If people want the “Instagram look” then they will have it. Are these pictures true photography? I believe that where a photographer has a vision for a photograph, then the image produced with or without the help of AI will be unique. Some photographers have an instantly recognisable style. Some embrace modern techniques, some insist that the only valid photography is created in a darkroom. All of these attitudes contribute towards the rich cacophony of modern photography.
To the second question, I say Yes, good photographers will always be in demand. At the beginning of my professional career I used to do a lot of food photography for not very much money. Now I do a little food photography for a great deal more money. For some of the same clients. Those clients don’t want a photograph that could be taken by anyone. They employ me because they know that I can produce the result they want. It is the lot of the professional photographer to constantly improve, to try new technology and integrate the good stuff into their work. This constant reinvention is a good thing, unless you’re a dinosaur!
So does AI improve photo processing? I think it’s an evolution, it’s a substantial improvement on filters and presets, it will make it possible for amateur photographers at the beginning of their journey to produce better pictures quickly. A revolution? No, by its very nature, building on existing knowledge, its not a revolution. The rule book is not about to be torn up. Rather, AI is a challenge and an opportunity.
If you enjoyed this post you may want to check out my earlier post – DxO Photolab Review or even check out the product – there’s a 30% discount until November 19th and while I do get a small percentage of the sales, this is a product I’m happy to endorse, You can download a trial version from the link below.