I recently became interested in the growing role that Artificial Intelligence, or AI, is already playing and will be playing in photography. Make no mistake, AI has already been deployed in our smart phones. And Adobe and a host of competitors have already enhanced their products with the first round of AI improvements. What effect will AI have on us, whether we be casual or serious photographers, amateurs or professionals? We will all be affected sooner or later. So, let’s take a look and see what’s happening right now and where AI is likely to take us.
There are two areas in which AI will have an impact. The first is in the camera itself. With AI, the camera will be able to anticipate challenges, especially in exposure and sharpness, and automatically set up the camera to avoid them. Today photographers must recognize these challenges and correct then for themselves.
The other area is in the post processing, what I like to call the Digital Darkroom. AI enhanced applications, such as Lightroom, will be able to look at a RAW image, analyze the type of image it is (landscape, portrait, etc.) and make optimum adjustments.
What is AI?
But before looking at the developments in AI, it is worth taking a moment to discuss exactly what AI is. As might be expected, there is some misinformation being disseminated about AI. Some are giving AI credit for everything a camera has ever done for the photographer, even features such as autofocus, auto-exposure and others that have been available long before AI was even possible. Virtually any decision the camera makes for us is labeled as AI by some. This is hardly the case.
AI, as far as photography is concerned, is part of a larger area of exploration and development called Computational Photography. A core process in this developing arena is Machine Learning. Probably the best example of machine learning, or ML, is IBM’s Watson, a computer that learned how to compete at Jeopardy when it went up against the top two all-time winners. There was no way the IBM team could give Watson access to the vast amount of facts it would need to know to even compete, much less win. One of the most challenging conditions was when the answer was ambiguous, something that humans were far better at figuring out. But giving Watson the ability to learn from its mistakes was the key to success. And Watson won.
ML, also called deep learning, is being applied to many aspects of our lives (think Seri, Google Now, Cortana, Alexa, etc.) and the players in photography are definitely making huge investments in it. Adobe is making a big play with Sensei, their ML initiative that will influence all of their Creative Cloud products.
Adobe is training its various AI algorithms on millions of images. This requires herculean computer resources, something that few companies can afford. But other would-be competitors are training their algorithms on tens of thousands and even hundreds of thousands of images. Theoretically, the more images, the better.
AI in Our Cameras
Apple, Samsung and Google are already incorporating AI in their smart phones. Because of the phones’ small formfactors, their cameras have pretty much reached their performance limits. Future improvements will come by way of AI features. One example that is already available is the Portrait mode that blurs the background. The camera is not capable of a shallow depth of field, so the background blur is achieved through AI.
DSLRs don’t have AI features (if you define AI as algorithms created from ML), at least not yet. But Arsenal is trying to fill that gap. It has a way to go but it’s on the right track.
Mirrorless cameras are better positioned to incorporate AI. There’s already talk about Sony integrating AI in their sensors (which, incidentally, they provide to Nikon) but that’s futuristic speculation at this time.
AI in the Digital Darkroom
AI is already making its way into the digital darkroom. Luminar purports to be a Lightroom alternative with built-in AI capabilities. It provides AI directed adjustments while still allowing the photographer to make their own. Photolemur is offered by the same company. It is specifically designed for the casual photographer who does not want to learn how to use a program like Lightroom. Photolemur only provides a one click solution, powered by AI.
Pixelmator Photo is available only on the iPad and offers several AI functions – ML Enhance (ML stands for Machine Learning) which applies adjustments to images and ML Crop which is advertised to even improve your composition.
Lightroom now contains two AI features. The first is Enhance Detail that is reported to increase detail in images by as much as 30%. Secondly, the Auto button in the Basic group in the Develop module that we all ignored is now improved with AI behind it. And it arguably does a better job than the old Auto button, making it an improved starting point for adjusting images.
Photoshop employs AI in an improved object selection function. The Select and Mask option in the Select menu can instantly identify and select an object that you click on. It removes the background. This is not such a great deal for photographers unless you like to create composites. But it’s a big deal for designers.
And Adobe is giving hints about using machine learning to learn the preferences of individuals. All they need to do is upload their works to the Adobe Cloud where they will be analyzed and used to make suggestions consistent with the individual’s preferences and styles.
The Potential Impact of AI
Where will AI take us? The future is just starting to unfold but here are some possibilities.
As far as AI in cameras goes, most people who take photographs will probably be unaware that it even exists. This will be the ubiquitous users of smart phones which already have AI-enabled features. More AI features will be added, they will be used, and they will produce better images. And most people still won’t know nor will they care.
But for those of us that shoot DSLRs and mirrorless cameras they will probably be welcomed, especially if they help us avoid mistakes such as forgetting to set our camera settings back after changing them for a special situation. It will be important, however, for us to be able to override the decisions the camera makes no matter how intelligent they may be.
As for the decisions our digital darkroom apps make for us, again, some will want the you-get-what-you-get approach as with Photolemur. They will be delighted because their photographs will be comparable to those created by professionals (at least in the digital darkroom part of the process).
Some professionals will love the AI enhancements that relieve them of hours of tedious masking, etc. that a product like Select and Mask in Photoshop enables. Wedding and portrait photographers will like the improvements in rendering skin tones that is already available and can only get better. In other words, a large body of professionals will become more productive with AI. And if Adobe makes good on the promise to be able to learn an individual’s preferences, the mark of individuality and personal style could even be retained.
How many, however, will accept the aesthetically pleasing results returned by AI apps and stop there? One of the marketing approaches promises photographers professional looking results without having to learn Lightroom or Photoshop.
But for those of us that see photography as a means of personal expression then AI could offer the starting point but not the final image.
It’s important to remember that a great photograph begins with fantastic light and a strong composition. Also, the exposure needs to be right on and the sharpness appropriate for the photographer’s intent. But this isn’t enough to create an image with impact. The image’s message is created or at least realized in the digital darkroom. And as long as the digital darkroom plays such an important role, those features that enable the photographer to achieve his or her goal with an image are the ones that matter, whether they are adjustments that have traditionally been available to us or algorithms that have been trained by machine learning. The important thing is that with expressive photography the human remains the essential ingredient.
Join me on one of my exciting workshops. Click here for more.
(320)
