Tag: AI

  • Is AI the future of Image Editing?

    Artificial intelligence has been a big thing in image editing in recent years. Adobe has had its Sensei for a few years, Topaz Labs has a whole set of AI-driven software, Skylum Software is launching Luminar AI in the near future. DxO Labs upgraded the Photolab to version 4 and added an improved AI-driven noise reduction.

    AI is a good thing.

    In general, I think AI is a good thing what it comes to image editing. It will make it easier and faster. I will get back to this, but before that, I will say a few words about noise reduction which I think is the best part of AI in image editing.

    Noise is something most photographers do not like. It can be a thing that ruins an image. Using a high ISO will introduce some noise. Also if an image is a bit under-exposed for some reason there will be noise. Especially if the underexposed image is brightened by increasing the exposure.

    Noise Reduction with AI

    AI-driven noise reduction is definitely a good thing. The quality has gotten better and better. AI can remove noise very well. I have used Topaz Labs DeNoise AI. I have been very happy with the results. If you want to hear more about what I think about Topaz Labs DeNoise AI watch this video. Also Adobe has AI-driven noise reduction in the making. It is a part of their Neural Filters which I will address a bit later.

    Just a few days ago DxO Labs launched version 4 of its DxO PhotoLab. It has three different noise reduction methods, HQ, PRIME, and the newest DeepPRIME. DxO PhotoLab 4 Elite version works as a plug-in for Lightroom. I have used DxO Photolab in the past. Now that the DeepPRIME noise reduction has been improved I will use it a bit more.

    I have not tested DxO PhotoLab 4 against Topaz DeNoise AI what comes to noise reduction. Both seem to work great and I will make a video where I compare them after I have tested DxO PhotoLab 4 a bit more. The first results indicate that it is better than Topaz DeNoise AI.

    AI can be used for a lot more than just noise reduction. Richard Harrington from Skylum Software said that Luminar AI will address a common problem. Quite a few do not know where to start when editing images. Luminar AI can give a good starting point for that. The same thing is with DxO PhotoLab 4. You can choose which correction you want the AI to do as default. Of course, everything can be adjusted, but they give you a starting point.

    Meet-Luminar-AI-2_1920.jpg



    I can understand that problem. I have been lecturing and teaching photography and image editing for many years. It is for many a real challenge to know what should they do with their image. Where to start. The main thing is first of course to understand how we look at images. What is the thing we see first in an image? Understanding that is very important. That will help you to edit.

    The problem with this is that it will give a general starting point based on what the AI has learned to be a good image. AI has based its view on millions of images. Eventually, this will lead to images that look all alike. AI thinks that landscape photographs should look a certain way.

    Will this affect our thinking about what is a good image. maybe it will. That has happened with Instagram. Those who use it most likely know what kind of images gets likes. I am not referring to those that show what I have eaten. Those likes come from close friends. I am talking about those images that have some photographic value. Images that can be appealing to a wider audience. If this happens with AI-driven image editing I do not think it is a good thing. This is a common problem with any technology. If it is there to help you to achieve the goals you have and in this case the photograph you like, then it is a good thing. If you let the AI decide for you then it is not totally your image.

    What about Adobe and Sensei?

    What about Adobe’s AI, Sensei how they call it. It has been quite helpful in video editing. A few examples are content-aware cropping and content-aware fill. In the latest version of After Effects, the Rotobrush 2 makes the green screen almost obsolete. In the latest version of Photoshop that was launch in Adobe Max has Neural Filters.

    Neural Filter includes Face Cleanup, Sky Replacement, Super Zoom, etc. Some of the Neural Filters are really useful. The Super Zoom lets you crop an image without losing any pixels. I did a quick test and it seems to work really well. The cropped version has a very good image quality. That is something that needs to be looked into some more. Sky Replacement is a good tool if you do a lot of composite images. It will make it so much faster to replace the sky. But in general use for your images, I do not think it is a tool that should be used. It is more for advertisement and art.

    Is AI the future of Image Editing?

    What about the question I asked in the title of this blogpost. Is AI the future of Image Editing? Simple answer, yes it is.

    What is the direction I would like the AI to go? As I wrote, now the way it is that AI learns from a vast amount of images. In the future, I would love to have AI that learns from my images, from my way of editing. It should learn what I like. One example is contrast, color vibrance, etc. AI should learn from my preferences on those and some other things. Then it would give me a better starting point for my edits.

    What do you think about AI-driven image editing? Good or bad?

    Disclaimer: Some of the links are affiliate links. If you click those and buy something I do get a small commission. You do not pay any extra.

  • Can AI replace photographers?

    Can AI replace photographers?

    Google Clips is small camera that uses AI to determine when to take pictures and in some cases about what. It is a small camera that can placed almost anywhere and it automatically takes photos of what it considers to be worth a photograph. Of course it does not move and it has to be placed. Clips then decide when to the photo. I understand that it likes movement. So if something moves in front of it, it snaps a photo. It also snaps a photo when a face comes close enough etc. It can recognise the faces that are those ones that are mostly saved on your smartphone from Clips. So it takes more photos of your kids than strangers. I have not used or seen Clip, but that is the basic concept I found out when I read about it. 


    By clicking the photo you will get directed to Amazon. It is a affiliate link and if you purchase the item I will get a small commision, but you will have the same price. Image: Amazon.


    The interesting question is: Can it replace a photographer? Without seeing the Clips in action I would say no. Even though AI is getting better and better it is not capable of making all the decisions that photographer makes when making photographs, yet. As photographer Ben Long said “The act of photography is the act of expression,” (quote from CNN:s article about Clips). So it would be an act of expression of the one that wrote the algorithm, not the photographers.

    Even though AI is not near? of getting to be on the level of a experienced photographer, it might be good enough for a common snapper. Clips might be able to take better photos than an average snapper. If placed correctly it can record images that might be good enough from a dinner party for example. Few Clips placed around the table might catch those images that can be used as memories from that party. Most likely it still cannot catch the decisive moment, but on the other hand how many snappers can?

    The problem with AI for now is that it can most likely take photos that are taken by the common “rules” of photography. But that is just the start from where a photographer begins to make the image. Very seldom photographs that follow every rule of photography are interesting. The interesting and “a good” photographs are the ones where the “rules” have been broken. There is some creative element in the photograph that the photographer has put in the photo. So far AI cannot do that. But we never know when that is going to happen. Most likely the next step is collabaration between AI and photographers to get better images.

    The link to the article from CNN that inspired me to blog about Clips and AI.


    disclaimer. The link to Google Clips is a Amazon affiliate link. If you purchase Clips I will get a commision, but the price for you is the same.           

  • Adobe is getting really serious with AI.

    Adobe is getting really serious with AI.

    Adobe introduced some new impressive AI driven features that are in the making. Adobe Sensei are the brains. As a photo and videographer the most impressive possible future features for me where of course the ones that had to with video and photography.

    Content Aware Fill is a great toll, but it does not work with every image and situation. Deep Fill is Content Aware Fill with steroids. It does not only analyze the surroundings, but it also tries to understand whats really underneath the element we try to fill. Another technology that tackles the same problem is Scene Stitch. I think its even more impressive than Deep Fill. Scene Stitch will replace a part of an image by going through millions of image from Adobe Stock. It uses part of an image that it finds from the stock and replaces the part that you have selected. Scene Stitch rises a few questions about copyrights and about the document value of a photograph. I understand that this is not a problem with design, illustrations, marketing and advertising. The problem might arise in those if it crosses documentary photography. It challenges the truth and the documentary value of a photograph. 

    Lets take an example that was pulled right out of Adobes YouTube-video about Scene Stitch.    
      

    Before (screenshot from Adobes YouTube-video)
    Before (screenshot from Adobes YouTube-video)
    On of the After photos that Scene Stitch suggested (screenshot from Adobes YouTube-video)
    On of the After photos that Scene Stitch suggested (screenshot from Adobes YouTube-video)

    Scene Stitch gives several suggestions. The user then can pick the one that works best. The tennis court is taken from a Adobe Stock photograph. The question that I think is that what kinda of a document is the after shot? The pond is replaced with tennis courts. There was no mentions in the presentation if someone gets paid when parts of the images will be used. I would assume that there will be money involved when stock image is used. At least I hope so.

    Of course these new features may or may not be part of future versions of Adobe Apps. Adobe is experimenting with different new things.

    Links to the videos in YouTube that I mentioned in the post:

    Scene Stitch

    Project Deep Fill
     

    The blogger is an Olympus European Visionary who´s native language is Finnish.