Some thoughts on AI processing and M43.....

D7k1

Mu-43 Hall of Famer
Joined
Nov 18, 2013
Messages
2,622
After playing with the new Neural Filters in photoshop and reading about the noise filter DeepPrime in DXO, I think that the 2 stop advantage of FF is going to disappear much faster than the camera industry thinks (which why I posted the post about making sure you vote for noise reduction in the beta section of the Neural Filters). It appears that all of the software companies know AI is the next battle ground. How much longer before subpixel shift is available from images for highres in these programs? High core counts and lots and lots of memory and large SSD's are becoming the norm giving the software folks F1 like performance for the cost of a "family" car?

Once Adobe gets AI noise reduction for me any thoughts (which are very limited) of FF will be gone. I don't expect my backup software On1 to be far behind on this. Where I live gets many, many rainy days and I spend a lot of time in Photoshop and Vegas PRO and just in the last 12 months or so I've seen AI creeping in at a faster rate, for me this is exciting. What do you think of the advance of AI and what it will mean for smaller sensor systems?

I've added a couple of Adobe Sky Replacement AI filter images. seconds vs much longer depending on how critical and hard the replacement is:


--
Editor's note: image added, Creative Commons CCO 1.0 licensed, author Wilfredor, from https://commons.wikimedia.org/wiki/File:Restoration_using_Artificial_intelligence.jpg
 

Attachments

Last edited:

BDR-529

Mu-43 Veteran
Joined
Jun 27, 2020
Messages
377
After playing with the new Neural Filters in photoshop and reading about the noise filter DeepPrime in DXO, I think that the 2 stop advantage of FF is going to disappear much faster than the camera industry thinks
There is only one minor flaw with this logic: the very same DxO4 and Topaz AI Denoise can be used to denoise RAW images taken with FF and presto! there's again 2 stop advantage.

Of course two full stops are in many cases enough to push low light M4/3 images into "good enough" category after which yet another two additional stops from FF are no longer needed.

I have posted few samples of m4/3 ISO 25600 shots which I processed with Topaz Denoise AI and they are perfectly usable as long as I don't have to crop them substantially but even so I was not able to shoot outdoor soccer game with f/4 zoom lens because fields around here have so poor lights that I was not able to push shutter speed below 1/120s.

Ergo: I still need FF and f/2.8 zoom lens and AI Denoise if I want to take really good action shots under conditions which will be normal here for the next 4-5 months.
 
Last edited:

D7k1

Mu-43 Hall of Famer
Joined
Nov 18, 2013
Messages
2,622
  • Thread Starter Thread Starter
  • #3
43 rumors talks about the Old Oly f2.0 zooms and then you'd be in the same ball park. For me it's all about lens size. If I could get 2 more stops from the PL100-400 I'd be very, very happy for wildlife. With the fast primes, heck even the Oly 300mm two stops is a major advancement especially with the size difference. Parity at the current level of FF would be a big deal for most folks I think.
 

grcolts

Mu-43 Veteran
Joined
Feb 1, 2010
Messages
472
Location
Texas
Real Name
Gary
Using faster lens and software seems to be the way to go when shooting in low light situations. I thought about picking up a FF for just low light shooting but for now I am quite pleased with my set-up. If future software will give me 2 stops advantage I would see no advantage to moving to a larger and heavier FF system. Time will tell.
 

mcasan

Mu-43 All-Pro
Joined
Feb 26, 2014
Messages
1,741
Location
Atlanta
Makes me wonder with mainstream camera bodies will add LiDAR focusing for low light and not depend totally on AF nodes on the sensor. If iPhone 12 can do it, no reason Canon, Nikon, Fuji, Panasonic and Olympus can't do it also.
 

BDR-529

Mu-43 Veteran
Joined
Jun 27, 2020
Messages
377
Makes me wonder with mainstream camera bodies will add LiDAR focusing for low light and not depend totally on AF nodes on the sensor. If iPhone 12 can do it, no reason Canon, Nikon, Fuji, Panasonic and Olympus can't do it also.
I belive that the main reason for LiDAR in iPhone is to support AI features like surprisingly credible "fast lens" DOF as a depth scanner. AI algorithms together with depth scanner enables also other nice things which cost hundreds if not thousands in ILC world like whatever bokeh you might ever want on any of your images.

All you need is one alpha channel with depth information and AI can modify DOF, bokeh and even focus point because they actually take several images when (and even before) you actually press shutter button. Practically nobody realizes that best smartphone pictures looks so good because they are actually AI enchanced focus stacked HRD images and most likely patched up from 2 or 3 different camera modules.

This is why ILC:s are losing this battle. They still try to capture one image as accurately as possible with this lens and these setting and let user do all the post processing if they so desire. In-body HRD is pretty much the only computational photography feature that actually works today.

Phone LiDAR has so little power that the range will be only few meters which is of course enough for a typical smarphone portrait i.e. someone sitting on the opposite side of the table and Apple does claim that low light AF becomes 6x faster and more accurate (which is to say that AF doesn't work at all withour LiDAR under very low light conditions). iPhone 12 has now even 5 axis IBIS so expect to see more and more smartphone users who don't need a separate mirrorless camera, not even one of FF variety.
 
Last edited:

Holoholo55

Mu-43 Hall of Famer
Joined
Aug 13, 2014
Messages
3,660
Location
Honolulu, HI
Real Name
Walter
I love post-processing software that reduces the noise and improves the images shot with M43 to level the playing field, so to speak, with larger sensor cameras. And, yet makes it easy for us to apply. I have no plans to get bigger cameras, so these improvements are welcome.
 

D7k1

Mu-43 Hall of Famer
Joined
Nov 18, 2013
Messages
2,622
  • Thread Starter Thread Starter
  • #9
Stop advantages can’t “disappear”. They are inherent in the physics.
Yes but we are talking about raising one level of technology to another through the use of computer power. It could be applied to the FF system for better performace, but as a bench mark we can say software can bring M43 upto (or close to) the performance level of FF. No one is saying that by increasing M43 we are decreasing the other technology.
 

Lawrence A.

Mu-43 All-Pro
Joined
Mar 14, 2012
Messages
1,723
Location
New Mexico
Real Name
Larry
I'm old. AI for me is mostly used in translation software, where I frequently refer to it as Artificial Idiocy, and Facebooks face recognition where it consistantly identifies me as someone else. I'm probably missing all sorts of photographic wonders, but I use my cameras much as I did when I shot film, though enjoying the better low light performance, "even with" M43, and the image stabilization. I do use noise reduction judiciously in Photoshop, because I mostly have it turned off in camera, and it doesn't really apply to raw files anyway. Reading this thread I realize how out of touch with current developments I am, and sad to say, unlike when I took up photography and read Ansel Adams' "The Negative" (which was my photo bible for years) I don't feel any compelling curiosity about much of it. I'm not suggesting others shouldn't. I'm sure the new technology will provide many advantages for all sorts of situations. All I'm saying is I pretty much have what I need, though I'd love a Leica CL digital and 23mm lens. Short of that, requiring the X 113 would suit me fine. I miss it. It's an old camera, but it gives me everything I need, even with its now measly 16 MP sensor, as does my Pen F and original E-M5. I had to look up LiDAR. It must be a very exciting time if you keep up with it and can use the new advantages. But though I love taking pictures, and seeing with a camera things I would otherwise overlook is important to me, the passion to study the latest technology is just no longer there.
 

D7k1

Mu-43 Hall of Famer
Joined
Nov 18, 2013
Messages
2,622
  • Thread Starter Thread Starter
  • #11
Hey, I am 71 so I am no youngster:), worked in the photo industry for about 15 years. Done commercial work in stills and video and had my images published in ads in quite a few magazines. Don't need the latest stuff but with M43 I find to get the best out of it software on a powerful computer is my tool of choice.
 

Lawrence A.

Mu-43 All-Pro
Joined
Mar 14, 2012
Messages
1,723
Location
New Mexico
Real Name
Larry
Hey, I am 71 so I am no youngster:), worked in the photo industry for about 15 years. Done commercial work in stills and video and had my images published in ads in quite a few magazines. Don't need the latest stuff but with M43 I find to get the best out of it software on a powerful computer is my tool of choice.
Well, I'm only 69, a mere kid 😁 but I did work in the industry - in labs as a custom printer working at enlargers - for 30 years, until digital caused all the local film labs to go belly up. I guess the passion for learning new photo skills just isn't there, though I understand why it would be for others. As a photographer my happiest times were working with a 4x5 view camera and printing my black and white work on fiber paper, tinkering with the chemistry and the process to get just the print I wanted. I knew what i was doing, and I was good at the craft, which always helps. I've never felt that kind of competence with digital, even though I know my way around Photoshop. I still like taking pictures, but that long, contemplative large format experience was the high point for me. Life isn't over but any means, but I no longer have a darkroom or a view camera and I suppose I'm more passionate about other things now, like learning a S.E. Asian language. I'm not a natural at learning languages, at all, so it takes a LOT of time. My previous post sounds like I'm throwing in the towel in life, when I'm just learning other things. Photography is a constant; learning new ways of doing things just doesn't have the WOW factor it once had for me, though I cannot imagine not taking pictures. I do work raw files and am still unwilling to turn my photos over to someone else for processing, and I just upgraded my desktop to handle music recording and photo editing more expeditiously, so I'm not a caveman, just getting a little starchy and set in my ways.
 

BDR-529

Mu-43 Veteran
Joined
Jun 27, 2020
Messages
377
But how much noise can be filtered out if the photos are already darn clean? So, the m4/3 has definitely much more to gain on the AI noise reduction issue.
FF owners are perfectly capable of pushing their gear instantly to new max level of acceptable noise if AI technology raises the bar.

Cost benefit for FF owners might be even higher because very fast FF pro/zoom lenses are insanely expensive. Instead of having to invest over $2000€ for say 70-200mm f/2.8 lens, same evening/indoor sport event could be shot with $700€ 70-200mm f/4-5.6.
 

D7k1

Mu-43 Hall of Famer
Joined
Nov 18, 2013
Messages
2,622
  • Thread Starter Thread Starter
  • #15
I am always impressed by how some 99% of the conversations are civil on this site, thanks to all who are participating in this.
 
Last edited:

bassman

Mu-43 All-Pro
Joined
Apr 22, 2013
Messages
1,225
Location
New Jersey
Real Name
The Bassman
1. “AI” is not a thing. It’s a marketing term for a complex set of computer programs. Or maybe not-so-complex programs. We’ll never know, since it’s the marketing guys that made it up. To paraphrase Arthur C. Clarke, any sufficiently advanced technology is indistinguishable from AI.

2. The impact of a performance boost for m43 vs. FF is dependent on whether the performance that is improved is already above the “acceptable” threshhold, whatever that might be. If (for arguments sake) the noise in a FF image is already undetectable, then reducing it by two stops adds no value. If the noise in an m43 image is acceptable for your use, then there is similarly no value added.

While I’m never one to turn down a performance improvement, m43 already delivers acceptable noise levels for nearly all of my images.
 

D7k1

Mu-43 Hall of Famer
Joined
Nov 18, 2013
Messages
2,622
  • Thread Starter Thread Starter
  • #18
1. “AI” is not a thing. It’s a marketing term for a complex set of computer programs. Or maybe not-so-complex programs. We’ll never know, since it’s the marketing guys that made it up. To paraphrase Arthur C. Clarke, any sufficiently advanced technology is indistinguishable from AI.

2. The impact of a performance boost for m43 vs. FF is dependent on whether the performance that is improved is already above the “acceptable” threshhold, whatever that might be. If (for arguments sake) the noise in a FF image is already undetectable, then reducing it by two stops adds no value. If the noise in an m43 image is acceptable for your use, then there is similarly no value added.

While I’m never one to turn down a performance improvement, m43 already delivers acceptable noise levels for nearly all of my images.
I think computer science people would disagree with your interpretation:


ar·ti·fi·cial in·tel·li·gence
/ˌärdəˈfiSHəl inˈteləjəns/

noun
noun: artificial intelligence; noun: AI
  1. the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages. I would say things like replacing a sky would or the ability to change the way a person looks requires decision-making.
 
Joined
Dec 6, 2015
Messages
1,199
Location
Brisbane, Australia
Real Name
Angus
I think computer science people would disagree with your interpretation:


ar·ti·fi·cial in·tel·li·gence
/ˌärdəˈfiSHəl inˈteləjəns/

noun
noun: artificial intelligence; noun: AI
  1. the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages. I would say things like replacing a sky would or the ability to change the way a person looks requires decision-making.
I'll be honest, I hit quote and then decided to walk away from it.

Sky replacements are small change. The true potentials of AI are huge.

AI is still new, but AI is already becoming more accurate at making medical diagnosis than doctors. Imagine where it'll be in 10 or 20 years. https://www.medicalnewstoday.com/ar...cally, the analysis found,with humans' at 91%.

The reason we won't fully understand AI is because AI and computer learning will become faster and more efficent than the rate we can understand it.
 
Links on this page may be to our affiliates. Sales through affiliate links may benefit this site.
Mu-43 is a fan site and not associated with Olympus, Panasonic, or other manufacturers mentioned on this site.
Forum post reactions by Twemoji: https://github.com/twitter/twemoji
Copyright © 2009-2019 Amin Forums, LLC
Top Bottom