Any good experience with AI tools to help organize or cull image

[RichF]

With all the advances in AI tools, I am wondering if anyone has found useful AI tools to help with the image library.

For example, a no-brainer would be tool which would step through your library and identify images that could merged into an HDR. Look for images that are bracketed and overlap.

LIkewise identify images that have similar subjects. Or have a subject in one and in another the subject is cropped. Or is not sharp.

It would be a great time saver.

 

[EricBowles]

I've run a couple of tests and had pretty good results. I found for things like wildlife, the results were not as good. I used Zenfolio Photo Refine. It did a good job with event photography on Day 4 of the PGA Tour Championship during an initial test. This test included awards photos where I had lots of groups of people.

The AI options are very good at eliminating out of focus images. But the software's idea of in focus does not always mean the image is truly sharp.

With people, it does a good job of selecting images with both eyes open - and the better images out of a group. But it may not know which subject is the critical subject for a group picture. So you could have the bridal party all included and the bride has her eyes shut.

Composition was not really a criteria.

Wildlife head, wing, and leg position was largely not considered. For me, this is a major reason for culling wildlife photos. With burst shooting I have far too many images that still need culling.

I expect to see a lot of development using AI tools. For example, I would pay for a product that added keywords based on bird identification. For the most part, these tools need to get a lot better to be truly valuable.

 

[Calson]

AI will have its place as machine learning and rules engines have been around for the past 30 years and useful in very limited and well defined situations. Want to optically sort fruit coming by on a conveyor line the application is useful.

It takes me very little time to cull images and for me the trick is to look at an image and in a second decide if it should be kept. I do not need to evaluate the exact reason why an image does not warrant keeping as it really does not matter.

I use software that allows me to view 4 image at the same time with each one taking up a quarter of the display. I find that I can decide which in a sequence to discard at this amount of enlargement.

 

[EricBowles]

AI will have its place as machine learning and rules engines have been around for the past 30 years and useful in very limited and well defined situations. Want to optically sort fruit coming by on a conveyor line the application is useful.

It takes me very little time to cull images and for me the trick is to look at an image and in a second decide if it should be kept. I do not need to evaluate the exact reason why an image does not warrant keeping as it really does not matter.

I use software that allows me to view 4 image at the same time with each one taking up a quarter of the display. I find that I can decide which in a sequence to discard at this amount of enlargement.

All that may still be true with AI. Using the Zenfolio AI tool, I was able to eliminate the likely discards and spend my time only on viable images for consideration. I find the role of AI is not to find the best images, but to help me avoid spending any time with the images that had clear flaws.

Even then, there may be some keepers that are addressed with current or future software. For example, if I have four images of a bride, it would pick the one with both eyes open and sharp, but its easy enough to replace an eye that is closed so I might prefer that photo due to the expression or other aspects.

 

[Nimi]

I've run a couple of tests and had pretty good results. I found for things like wildlife, the results were not as good. I used Zenfolio Photo Refine. It did a good job with event photography on Day 4 of the PGA Tour Championship during an initial test. This test included awards photos where I had lots of groups of people.

The AI options are very good at eliminating out of focus images. But the software's idea of in focus does not always mean the image is truly sharp.

With people, it does a good job of selecting images with both eyes open - and the better images out of a group. But it may not know which subject is the critical subject for a group picture. So you could have the bridal party all included and the bride has her eyes shut.

Composition was not really a criteria.

Wildlife head, wing, and leg position was largely not considered. For me, this is a major reason for culling wildlife photos. With burst shooting I have far too many images that still need culling.

I expect to see a lot of development using AI tools. For example, I would pay for a product that added keywords based on bird identification. For the most part, these tools need to get a lot better to be truly valuable.

Given their business model, any idea whether they harvest members' images to feed their AI models?

 

[EricBowles]

Given their business model, any idea whether they harvest members' images to feed their AI models?

I don't believe they do. They don't make money on this AI tool. It is from a third party. I suspect they believe it is part of the value they provide - especially for event and wedding photographers which make up a large portion of the business. They don't create any content with the AI - they just identify good and bad images based on criteria.

 

[Nimi]

I don't believe they do. They don't make money on this AI tool. It is from a third party. I suspect they believe it is part of the value they provide - especially for event and wedding photographers which make up a large portion of the business. They don't create any content with the AI - they just identify good and bad images based on criteria.

Assuming it's LLM AI, the model had to be fed photos classified as good or bad at some point, and they are storing images for their main business, and might injest images that they process. Maybe they can identify eyes and faces with just rules.

If it's important to you, I'd try to find out whether they're accessing the images you edit (and others in your portfolio depending what access you granted them) to be sure. Their website explanation is light on details and doesn't (that I can find) preclude it, nor (as Adobe Stock does, for example) give the user an opt-out.

 

[EricBowles]

Assuming it's LLM AI, the model had to be fed photos classified as good or bad at some point, and they are storing images for their main business, and might injest images that they process. Maybe they can identify eyes and faces with just rules.

If it's important to you, I'd try to find out whether they're accessing the images you edit (and others in your portfolio depending what access you granted them) to be sure. Their website explanation is light on details and doesn't (that I can find) preclude it, nor (as Adobe Stock does, for example) give the user an opt-out.

It doesn't make any difference to me. It would be one thing if they were using images to create content, but that's not what this is.

 

[Johnswarn]

Sure! There are AI tools like Adobe Sensei and Google Photos that can help manage image libraries. They organize photos automatically based on what's in them, like faces or objects. Clarifai and Imagga also use AI to add descriptive labels to images, making them easier to find. These tools can't automatically merge images into HDR or fix blurry pictures, but they're still helpful for keeping your photo collection neat and tidy.

 

[apanella]

For me, AI as a tool to increase my workflow efficiency is a noble goal.
(I am not yet sure about AI for image manipulation. So far, it is not for me)


PhotoSupreme DAM, from their website. https://www.idimager.com/
"Photo Supreme offers integrated Face Recognition capabilities and integrates with Google Vision to bring the most advanced technology available today to your desktop. Use Artificial Intelligence to help you jump start your cataloging by automatic labeling, object recognition, landmark recognition and face detection."

I have used the tool for automatic Keywording, landmark detection, and some (not much) face detection.
Seems to work OK. Takes advantage of the Google API for AI keywording. It takes around 1000-2000 photos to pass the Google API "per month" free tier.

This being said, keywording is not for the slight of heart. Be prepared for some work to get the AI keywords in agreement with pre-existing keywords.

The same tool can stack and/or version files based on rule sets (not AI) which is handy for post processing of image groups.


CaptureOne does some AI to identify eyes/faces as part of their culling module. Seems to work for people, but I don't photograph to many people.