Capturing Beauty, Made Easier: Can AI Lend a Hand in Your Photography?

Discussions
1

As members of the Nature Photographers Network, we share a common bond – a deep appreciation for the natural world and the desire to capture its beauty through our lenses. In this journey, we continually seek ways to refine our craft. One emerging tool that’s stirring conversation is Artificial Intelligence (AI). But how does it fit into the traditional art of nature photography? And more importantly, can it enhance our work without diminishing the value of our skills and the community’s insights?

AI: A Supportive Tool, Not a Replacement

The idea of integrating AI into photography might initially seem daunting or even unwelcome. However, it’s important to view AI as a supporting tool rather than a replacement for our artistic vision and technical skills. AI in photography primarily offers analytical assistance – it can help us identify areas in our images that might benefit from slight adjustments, like exposure or composition.

Acknowledging the Double-Edged Sword of AI in Photography

There’s an understandable concern among photographers about the rise of AI-generated images. These tools, while impressive, can create works that challenge our notions of authenticity in art. This concern is particularly acute for us as nature photographers, where the authenticity of the moment captured is paramount. However, it’s important to distinguish between AI used for generating images and AI used as an analytical tool. While the former can sometimes blur the lines of artistic originality, the latter serves to enhance our capabilities without infringing on our creative integrity. When used judiciously, AI can aid in aspects such as image enhancement and editing workflows, saving time and allowing us to focus more on the creative and interpretive aspects of our photography. It’s about leveraging AI to support our vision, not replace it.

Complementing Human Insight

The heart of photography, especially nature photography, lies in capturing a moment that resonates emotionally and visually. This is where the human element – our artistic eye and emotional connection to the scene – is irreplaceable. The critiques and discussions within our community are invaluable, offering perspectives and insights that no AI can replicate. AI can, however, provide a different kind of feedback, one based on technical analysis, which can complement the rich, subjective feedback we receive from our fellow photographers.

Encouraging Exploration and Questions

I encourage you, my fellow enthusiasts, to approach AI with an open mind and a sense of curiosity. How might AI highlight aspects of your work you haven’t noticed? In what ways can it streamline the technical side of your photography, giving you more time to focus on the creative aspects? These are questions worth exploring.

Balancing Technology with Tradition

As we navigate this potential new tool, it’s crucial to strike a balance. We must ensure that the use of AI aligns with our core values as photographers who cherish the authenticity of the natural world. AI should aid our creativity, not overtake it.

Understanding AI’s Limitations: Technical Flaws vs. Personal Aesthetics

While AI can proficiently identify technical aspects like exposure errors or compositional imbalances, it’s important to recognize its limitations in understanding personal aesthetic values. For instance, AI might suggest brightening a shadowy area in a landscape photo. However, as the photographer, you might have intentionally captured the shadows to create a mood or highlight a certain element. This is where the human element triumphs. Your artistic choice, driven by personal style and emotional intent, is something AI cannot comprehend or influence. It’s this unique personal touch that makes each of our photographs not just technically proficient, but also emotionally powerful and artistically distinct.

Conclusion: A Step Towards Enhanced Creativity

In conclusion, AI in photography isn’t about diminishing our art or the value of our community’s insights. Rather, it’s about enhancing our ability to capture the natural world in all its splendor. As we continue to share, learn, and grow within this vibrant community, let’s remain open to the possibilities that AI might offer. It’s not about replacing what we have; it’s about adding another layer to our ever-evolving art form.

2 Likes
2

Don, thanks for this article. Very intriguing.

I’d like to start out by saying that we (everyone) are misusing the term “Artificial Intelligence.” AI in its truest form is that point when a machine gains self-awareness…it’s called a singularity in scientific circles. What we have today is more specifically “Machine Learning.” Machines are instructed to collect information from diverse topics and codify that information to make top choice recommendations for the task at hand.

This is what is happening in photography today. The latest noise reduction techniques in Adobe and DXO, for example, rely on thousands or even millions of images to determine the best way to clean up an image. I use Capture One and it has a new technology called AI Masking. The program has compiled thousands of example images so it can select the subject in a picture. It can select the background; easy since it’s the opposite of the subject. It can also select individual parts of the image for a mask. But—and this is so important—Capture One is not self-aware. It’s only doing what the developers instructed it to do.

Adobe also has Generative Fill. Same Machine Learning concept coded by the developers. I’m sure Capture One will have that soon too. And I am sure that both tools will have new Machine Learning technology soon. Personally, I hope that one of these tools will have a way of applying color cast correction because I am red/green color blind. I would do anything for that.

I use DXO for noise reduction and Capture One’s “AI Masking” all the time, they’re game changers for my photography. I can take an ISO 6400 image and know that DXO will clean it up for me. I know that I can easily adjust the subject separately from the background of an image in Capture One.

So, thanks again for your article. I, for one, am embracing Machine Learning. However, I am concerned about the potential for Self-Aware Machines. I’m reminded of “The Terminator.”

“The machines decided our fate in a microsecond.”

3 Likes
3

Hi David thanks for your response and for highlighting the important distinction between ‘Artificial Intelligence’ and ‘Machine Learning’ – a distinction that indeed merits clarification, especially in the context of photography.

You’re absolutely right. The term ‘AI’ has become a catch-all phrase that often overshadows the more accurate description of what we’re dealing with: Machine Learning (ML). ML, as you’ve adeptly described, involves algorithms learning from vast amounts of data to perform specific tasks, like noise reduction or subject selection in photos. This is quite different from the concept of a self-aware AI, a singularity that remains in the realm of science fiction and theoretical science.

Your examples of Adobe’s noise reduction techniques and Capture One’s AI Masking vividly illustrate how ML has become an invaluable tool in our workflow. These technologies are indeed not self-aware; they’re sophisticated tools designed to process data and make recommendations based on patterns they’ve ‘learned’ from a vast array of examples.

Your anticipation for future ML developments, particularly in the realm of color cast correction, is something many of us share. The potential for these tools to adapt and address specific needs, like your mention of color blindness, is genuinely exciting. It opens doors to making photography more accessible and enjoyable for everyone.

Moreover, your point brings to the forefront another crucial aspect – the importance of understanding these technologies. As photographers, especially those of us embracing the natural world, staying abreast of technological advancements is not just about maximizing the potential of these tools. It’s also about maintaining relevance in an ever-changing landscape. In an era where technology evolves at breakneck speed, a solid grasp of Machine Learning and its applications in photography can empower us to make informed decisions about our craft. This knowledge allows us to thoughtfully integrate new tools into our workflow, ensuring that we continue to produce work that resonates with authenticity, even as we navigate the tides of technological change.

Your concerns about the potential for self-aware machines are understandable and shared by many in the scientific community. While we’re currently far from that reality, it’s a subject that merits ongoing ethical discussion as technology continues to advance. For now, ML remains a tool – a remarkably advanced one, but a tool nonetheless, under our control and designed to enhance our creative capabilities.

On a personal note, I’ve recently ventured into creating and using a customized GPT model, tailored specifically to critique my work from a technical standpoint. This journey with Chat GPT has been quite enlightening. It offers critiques (sometimes brutal) based on technical parameters, providing an additional, objective perspective that complements my artistic vision. However, what’s key here is how I use this feedback. I view the ML’s suggestions as a guide, not a directive. It’s a delicate balance, considering the technical insights provided by the ML, while ultimately letting my own artistic vision for the image prevail. This approach ensures that while I’m open to the technical improvements suggested by ML, I never compromise on the unique aesthetic that defines my work. It’s about harnessing technology to enhance creativity, not override it.

David, your “Terminator” reference is spot on – though I must say, if my ML ever starts muttering about ‘Judgment Day,’ I’ll promptly remind it that its biggest judgment call should be about exposure levels, not world domination. At least for now, the most formidable uprising we face is when the ML mistakenly over-sharpens a sunset!

1 Like
4

Thanks for the response, Don. I fully agree with all your comments. Interesting about the customized GPT model. I may have to figure that out soon.

Cheers,
David

5

Let me know if you need any pointers on the GPT always happy to help. :+1:

6

Very interesting discussion, and much more nuanced than I expected. I was distressed to find on a first – and last – look that something I mistook for a more sophisticated content-aware fill offered to put a red car of my choice on the top of a slightly cropped tree. I sort of prefer the old way of learning to critique my images, slow and humiliating as it may be.

I have nothing to add now but I’ll try to keep up with future posts. Thanks for the thoughts and information!

1 Like
7

Don,

Nicely written. I agree with you to the extent this relates to the photographer’s creative and/or basic processing options and tools. With the main premise the original image is OUR captured image; it’s MY original creation - whether it’s film, digital, your iPhone, whatever. WE created the image. The image being processed was NOT created by AI. With that said, I support “editing” my creative works with the available tools. Specifically, I’m pretty impressed with the recent addition to Photoshop of the AI based “Remove Tool” and the related tools in LR/ACR, Content-aware fill, etc. Pretty amazing how cleanly items can be removed. (Adding non-existent items, I do not support.)

I don’t have any issues removing an errant branch, a mis-placed rock in a stream, a blown-out piece of sky in a tree canopy, etc. etc. - removing a beer can in a meadow… hopefully all understand the limits to what I’m talking about. Using a transform tool to “bend” a tree or building… I’m good with all that. And if AI is employed for that purpose, I’m good with that.

I expect and assume this is at the core of your writing. Being “open” to the use of the tools as they are to, my words: “making the image the best it can be.” The assumption being the original image is yours.

I do NOT support using these same tools to make an image something that it wasn’t. Which leads to my significant concern of this whole AI thing. Big picture, I know there are monumental potentials with AI when it comes to science, cures for disease, etc. etc. On the flip side, there are monumental and real risks for catastrophic harm… but I digress and will keep this to photography.

My biggest concern - and this is already bearing out - is that we no longer can trust or believe an image is originally created by a person, OR created by AI. I’m responding to this post because I just recently experienced this on a post on FB. A image popped up in my feed “Breathtaking shot - Yosemite National Park.” Currently 445 comments, the overwhelming majority, “Beautiful!”, “OMG!” blah, blah, blah. A small handful, including myself called this out as an AI creation. But the mindless masses likely think it’s a real photograph, which are beared out in the hundreds of comments. That my fellow NPNer’s is the danger and reality we are facing with AI/ML. I’ll refrain from posting the name of the “photographer” who posted this… but in reviewing their website, I have zero confidence to believe any of their work is real. and the sad part is, maybe much of their original photography is real. But this image, tarnishes everything. This is the future, and I’m quite disturbed by it. Here’s the image posted on FB just a week or so ago:

image
image641×967 193 KB

There is so much terribly wrong with this image, but the main thing is the blocks to the right of “El Capitan” (btw El Cap is hideous in this image.) Only someone familiiar with Yosemite would know this view/perspective doesn’t exist… and so I knew immediately this was AI. Many others called this out as well. Sadly, this is the reality of today as far as photography on social media goes. But also AI creations have already been exposed on juried photo competitions - print and online. Extrapolate this to the rest of humanity… and we’re in trouble. (my apologies, I tried hard to keep this about photography - so please ignore if you can.) :slight_smile:

Regarding AI critiques… let us consider how AI works. AI - or as David clarifies, ML, works by culling thousands and actually millions of “critiques” farmed from every conceivable online resource. Let’s be honest here - the overwhelming majority of these critiques will be sourced from Instagram and Facebook. Not NPN. SO, the AI “critique” will be primarily based on what is formulated on these social media platforms. I personally have zero desire to get feedback, artificial or direct, from Social media users. But that is how AI/ML is currently working. Of course it’s more complicated than that when it comes to technical critiques like exposure, sharpness, etc. etc., but by in large, AI by definition is learning and consolidating from millions of “opinions.” No thanks. I want YOUR opinion, your critique, Diane’s opinion, Diane’s critique… opinions from real people.

So yeah, it’s good to have an “open mind” on something as significant as the subject of AI as it relates to processing our images and the tools we use to edit/create our photographic works… but I would say it’s even more important to critically think of its broad application and ramifications. Tread lightly folks, and THINK. Clone a beer can, fine. Create something that never happened… not so much.

2 Likes
8

Funny, my last blog post talks about this in terms of how connected to my work I feel and how removing or adding too much distances me from it. Overall that’s not good and so my measure of how far is too far is based on that connection.

1 Like
9

How far is too far is a subjective evaluation. That means that some degree of AI is good sometimes but there can’t be full agreement on when that is.

It seems to me that the cat was let out of the bag when PS introduced the clone stamp tool. We say that removing unwanted objects is fine but adding new object is crossing the line. However, cloning is not removal. It’s replacement. And replacement is inserting new information. I suppose inserting information from the very image is acceptable but inserting it from other unrelated images is not. That’s a slippery slope in my opinion.

Then there’s the argument ‘The proof of the pudding is in the eating’. That argument supports the idea of judging the result and not the process. It’s like the desk you bought. Is it made from real wood or composite wood glued together. Many people buy it for how it looks and not for its structural components. I actually don’t like AI at all but I’m playing the devils advocate here. I suspect that AI imagery will soon grace the living rooms of most Americans (much like IKEA furniture).

3 Likes
10

Hi Lon thanks for your thoughtful input. I want to start by clarifying that my exploration into Machine Learning (ML) for critiquing my photography work doesn’t stem from a desire to push this technology onto others as a one-size-fits-all solution. Like any tool in our photographic arsenal, from the most basic to the most sophisticated, its value is realized in how we, as artists, choose to wield it. My journey with ML is born from a belief in its potential to enhance our work, much like how an expensive camera can elevate our artistry when we fully grasp its capabilities and limitations. This belief is grounded in the idea that technology, when thoughtfully integrated, can serve to expand the boundaries of our creative expression without compromising the authenticity at the heart of our craft.

The distinction between analytical AI tools, like the custom ML model I’ve developed, and image generation technologies such as Midjourney and DALL·E, is crucial in this discussion. My model is designed to provide technical feedback based on the fundamentals of good photographic practices, offering an unbiased, unemotional critique that identifies areas for potential improvement. It draws on principles of photography rather than trends or preferences culled from social media, ensuring that the feedback it provides is relevant and grounded in enhancing the technical quality of the image without dictating its artistic direction. This approach allows me to take or leave its feedback based on how it aligns with my vision for the image, ensuring that the final outcome remains unequivocally mine.

I share your concerns about the implications of AI in image generation, particularly the challenges it poses to authenticity and trust in the photographic community. The example you cited of the Yosemite image is a poignant illustration of the broader ethical and philosophical questions we must navigate as these technologies become more pervasive. As photographers, we are custodians of the moments we capture, tasked with presenting these slices of reality in a way that is true to our artistic vision while respecting the truth of the scene before us. The advent of AI-generated images complicates this responsibility, blurring the lines between reality and fabrication in ways that can undermine the trust placed in us by our viewers.

Your point about AI critiques, and their reliance on a vast, impersonal pool of data predominantly sourced from social media, further underscores the need for a discerning approach to integrating these technologies into our workflow. The value of personal, experienced-based critique cannot be overstated, offering insights and perspectives that are deeply rooted in the human experience of both creating and viewing art. It’s a reminder that while AI can serve as a tool for technical refinement, it cannot replace the nuanced understanding and emotional resonance that come from genuine human engagement with our work.

In embracing AI, it’s imperative that we do so with a critical mind, recognizing both its potential and its limitations. Like any tool, its effectiveness is contingent upon our ability to integrate it into our creative process in a way that enhances rather than diminishes our art. This requires an ongoing dialogue within our community, where we share our experiences, challenge our assumptions, and collectively navigate the evolving landscape of photography in the age of AI.

Thank you again for your insights, Lon. Conversations like this are invaluable as we strive to understand and adapt to these changes, ensuring that our art continues to resonate with authenticity and integrity.

1 Like
11

Girrakool Loop-38-Edit-2
Girrakool Loop-38-Edit-28417×4562 1.9 MB

When embarking on the post-processing journey of any given image, I often turn to my custom Machine Learning (ML) model for an initial critique. This tool isn’t there to dictate the artistic path I choose but to serve as an analytical first glance, offering feedback that I may consider or disregard according to my creative vision. With each piece, like the serene waterfall scene captured here, the model provides a ‘Technical Facet Scorecard’—a take-it-or-leave-it breakdown that helps me reassess the image’s elements before I delve into the editing process.

Here is a sample of the feedback it generates:

Introduction

This photograph presents a serene waterfall nestled within a dense forest. The natural framing with the intertwining branches suggests an intent to highlight the waterfall as the central subject, while also showcasing the surrounding foliage and the rugged terrain.

Technical Facet Scorecard

Composition: 7/10

  • The branches form a natural arch leading the viewer’s eye toward the waterfall, creating a strong focal point.
  • Reason for Loss of Points: The left side of the frame is slightly heavy due to the dense foliage, which slightly unbalances the image.

Exposure: 8/10

  • The exposure is well-managed, preserving details in both highlights and shadows.
  • Reason for Loss of Points: There’s a hint of overexposure in the water’s highlights, losing some detail.

Focus and Sharpness: 9/10

  • The image is generally sharp, with the main elements in focus.
  • Reason for Loss of Points: Some foreground elements appear slightly soft, which may be due to the depth of field used.

Color and Tone: 7/10

  • The color palette is natural and appropriate for the scene.
  • Reason for Loss of Points: The colors could be more vibrant; the image seems a bit flat, lacking contrast.

Lighting: 7/10

  • The diffuse lighting minimizes harsh shadows and highlights the texture of the forest and water.
  • Reason for Loss of Points: The lighting is flat, which diminishes the sense of depth.

Overall Score: 7.6/10

Score Matrix:

Composition Exposure Focus and Sharpness Color and Tone Lighting Overall
7/10 8/10 9/10 7/10 7/10 7.6/10

Suggestions for Improvement

In-Field Adjustments:

  • Experiment with different angles to potentially find a more balanced composition.
  • Use a polarizing filter to reduce reflections and enhance the colors.

Post-Processing Tips:

  • Increase contrast and saturation slightly to bring out the colors and depth.
  • Apply selective sharpening to the foreground to enhance detail.

Additional Insights

The mood of the photograph is tranquil and reflective, inviting viewers to appreciate the quiet beauty of nature. The use of natural framing is a creative approach that adheres to the ‘frame within a frame’ concept, drawing the viewer’s eye directly to the waterfall.

My model operates on a set of principles that underpin traditional photographic excellence while integrating the capabilities of machine learning for technical analysis.

Here are the detailed principles:

  1. Composition: The model evaluates the arrangement of elements within the frame, considering classic rules such as the rule of thirds, leading lines, symmetry, patterns, and the use of negative space. It assesses how these elements guide the viewer’s eye and whether they contribute to or detract from the main subject.
  2. Exposure: It analyses the luminance values to determine if the photo has a balanced exposure. The AI checks for clipped highlights, blocked shadows, and overall tonal range, providing feedback on how well the image captures the scene’s dynamic range.
  3. Focus and Sharpness: Clarity is crucial in photography, and the model assesses the focus accuracy on the subject, depth of field appropriateness, and the overall sharpness of the image. It also looks for any unintended motion blur or camera shake.
  4. Color and Tone: The AI reviews color balance, saturation, and vibrancy. It considers whether the colors are true to the scene and if they evoke the intended mood and atmosphere. It also checks for color casts or unnatural skin tones where relevant.
  5. Lighting: Understanding that lighting is the cornerstone of photography, the model critiques the quality, direction, and softness of light in the image. It evaluates how light and shadow are used to create depth, dimension, and emphasis.
  6. Authenticity: In line with your mantra, “The Authenticity of Nature,” the AI is tailored to appreciate the natural beauty without overemphasis on digital manipulation. It respects the integrity of the original scene.
  7. Technical Precision vs. Artistic Intent: While the AI can spot technical flaws, it’s designed to understand that not all technical ‘imperfections’ are unintentional. Sometimes, creative choices deliberately flout technical norms to achieve a particular artistic effect.
  8. Feedback and Suggestions: Beyond merely scoring, the model offers constructive feedback and potential solutions. It provides actionable advice for both in-field adjustments and post-processing enhancements.
  9. Customizability and Learning: Recognising that photography is subjective, your model can be fine-tuned to align with personal aesthetic preferences or specific genres of photography.
  10. Ethical Use of AI: The AI is built with the understanding that it should not encourage or suggest manipulations that would deceive viewers or misrepresent reality, especially in the context of nature and documentary photography.
12

Thanks Igor.

You raise an intriguing point about the inherent subjectivity in determining the limits of ML’s role in photography. Indeed, there is no universal consensus on when the use of ML transitions from being a helpful tool to one that potentially oversteps the mark.

The introduction of Photoshop’s clone stamp tool did indeed mark a significant moment in our ability to manipulate images. Your perspective on the act of removal being akin to replacement is a valid interpretation and adds to the complexity of defining ‘authenticity’ in digital imagery.

Your ‘slippery slope’ argument presents a valid concern about where we draw the line. While some might find the replacement of pixels within an image acceptable if it enhances the final result, others argue that the integrity of the photograph is compromised the moment any pixel is altered, regardless of the source.

The ‘proof of the pudding’ analogy is another compelling argument. It reflects a sentiment where the end justifies the means, a viewpoint that resonates with many in our visual-driven culture. It’s clear that as ML continues to develop, the aesthetic value of images may increasingly take precedence over the process for the average consumer, much like the appeal of a well-designed piece of furniture might overshadow its material composition.

While playing the devil’s advocate, you’ve touched on a key future trend: the likely ubiquity of ML in our visual spaces, akin to the prevalence of IKEA furniture. It’s a thought-provoking parallel that challenges us to think about the value we place on the ‘handcrafted’ versus the ‘machine-generated’.

Thank you for your candidness, Igor. It’s through such challenging dialogues that we can better understand the spectrum of opinions within our community and reflect on our individual practices.

1 Like
13

Agree with this Igor. You’re right, we’re still “modifying” the original scene.

If I were to simplify my personal position on this, it would be this. If I could have walked in to the scene and removed the beer can, then I have no problem doing it with a clone-removal after the fact. I don’t really care if the replacement of the blades of grass came from another part of the image, or from a thousand composites, the point is I removed the beer can. But I do have a problem inserting a bugling elk or crescent moon in to the scene that was never there. Replacement, sure I agree.

My simple argument can be shredded easily by arguing that it’s not possible for me to go and bend a ponderosa tree trunk at the time of the original capture… :slight_smile: But hopefully my simple thought process makes sense.

I agree with this too. I’m trying to create what I saw and experienced. By adding or changing sometime too much that it no longer represents what I experienced, then I might draw the line.

Of course with ICM and abstracts, it all goes out the window!! ha ha. But seriously, I just thought of a question that could be requested from Chat GPT: “Please create an intentional camera movement motion blur image of quaking aspen leaves in the fall” Then what? I’ll never do it, but pretty sure someone will…

1 Like
14

It seems we have veered into the broader and indeed contentious issue of creating or significantly altering images with AI. While that is a discussion worth having, I’d like to steer us back to the core topic at hand: the use of AI/ML as an analytical aid in photography. This is about using technology to refine our craft, not to fabricate or misrepresent our work. AI/ML, in this context, is a tool for enhancing our understanding of the technical aspects of our images, ensuring that the authenticity of our captures remains intact.

1 Like
15

Thanks Don for keeping this focused. I know I’m guilty of diverting a bit to the broader AI concerns, so thanks for gently bringing it back… :slight_smile:

Personally, I’m way beyond needing or wanting guidelines and virtual assistance after some 50 years behind the viewfinder… but I certainly applaud your efforts in developing learning tools.

I think there is great value for those who are learning and want to develop! Don’t get me wrong, one is never to old to learn and I’m certainly not opposed to feedback and learning - heck, that’s one of the reasons I’ve been on NPN so long!

Charge ahead and good luck!

1 Like
16

WOW fifty years behind the viewfinder! —now that’s a milestone that certainly earns you the right to forego any juggling of virtual balls, AI, ML or otherwise! But just as you’ve said, we’re never too old … this old dog still wants to learn to juggle but then I know after fifty years behind a viewfinder I am likely to feel just the way you do. Can’t beat experience!

1 Like
17

Hey David. Your analysis is very thorough, it makes it easy for people to understand this knowledge, I am very grateful to you.