Luminous Landscape Forum
Raw & Post Processing, Printing => Digital Image Processing => Topic started by: plugsnpixels on August 16, 2018, 04:24:14 am
-
Release date: August 16.
A.I. Gigapixel is a new image resizing app contender, for both JPEG and RAW images. See my blog post with examples and details (http://plugsandpixels.com/blog/new-topaz-ai-gigapixel-upsampling-app-w-unadvertised-discount/). An unadvertised discount is available.
-
Release date: August 16.
A.I. Gigapixel is a new image resizing app contender, for both JPEG and RAW images. See my blog post with examples and details (http://plugsandpixels.com/blog/new-topaz-ai-gigapixel-upsampling-app-w-unadvertised-discount/). An unadvertised discount is available.
Hi,
Thanks for the heads up. I look forward to trying it. Unfortunately, I cannot log-in upon starting the program (it says it's "unable to connect to the server"), although I can access my Topaz account via a webbrowser with the same credentials.
Probably a V1.0 bug, or they are not yet finished setting up things, and I have to be a bit more patient.
Cheers,
Bart
P.S. And here is the official A.I. Gigapixel website:
https://topazlabs.com/ai-gigapixel/
-
It would be useful to see a comparison vs Photoshop CC Details 2 up-sampling, which does a fantastic job.
-
It would be useful to see a comparison vs Photoshop CC Details 2 up-sampling, which does a fantastic job.
+1 and ideally, in print (unless the tool is to work solely for screen output). Yeah, I know, that's hard to do but what can look butt ugly on-screen can look a lot better than the alternative on a print, assuming that's your goal.
-
Thanks all, I had begun doing the PS comparisons and will continue later today. Everyone can now give this a try (I revised my blog links to get you straight to the now-released download).
True about prints being the main output method of such technology.
-
In a few tests I have done, it appears that the AI Gigapixel algorithm handles JPEG artifacts around edges better than PS Preserve Detail v 2.0. Very limited testing though (a few images!) when upscaling 200%, default settings. Otherwise, the results were not much different than PS, other than the fact that the AI-GP upscaling took a few minutes (MacBook Pro, 2015 - AMD Radeon R9 M370X 2048 MB) and PS took a few seconds (original image size was 3000 x 2000 pixels, for example). See if this agrees with your experience when testing both applications.
Kirk
-
Thanks Kirk!
I updated my blog post with some A.I. Gigapixel vs. PS results.
-
Thank you!!! One tip to show the comparison to Photoshop a bit better, turn off Pixel Grid in Photoshop on the fountain comparison :)
-
Good point... ;-)
-
Here's another 1-megapixel sample (shot at Cal Tech in the late '90s, plus full-frame reduced size for reference), enlarged to 400%, Photoshop on top (Preserve Details 2.0 + 10% noise reduction), Topaz on bottom (with Enhance Image checked).
-
Here's another 1-megapixel sample (shot at Cal Tech in the late '90s, plus full-frame reduced size for reference), enlarged to 400%, Photoshop on top (Preserve Details 2.0 + 10% noise reduction), Topaz on bottom (with Enhance Image checked).
Thanks
That’s actually quite impressive.
Cheers,
Bernard
-
Time to process is very long, around 20 minutes to uprez a D850 image to 30 x 40 at 300 dpi or another run just uprezing to 150%. On my Macpro, 1 year old, 4 way, 64GB of ram High Sierra.
Comparing at 100% to same file run on Adobe CC "enlargement and preserving details", images appear very similar, thus for me hard to justify the 20 minute run time.
Topaz also seems to have no way to keep the original file name, (at least I can't figure it out) as the file name is truncated to about 25 characters. You have an option for a "output" suffix, so maybe you could rewrite the full file name, but that's not what I would want to do.
Tool may have better results on jpgs or smaller files, have yet to try anything besides D850 images.
Paul C
-
Thanks Bernard and Paul. If I can today, I will test on my Hackintosh and see how the speed is.
I suspect the product will be further optimized over time.
-
I would appreciate your opinion on a 24MP to 40MP file uprezed to just 150%.
I may have a setting wrong, but the tool takes a very long time on my Mac's. I will try in on Windows later today. And on a few smaller 20Mp Drone images.
The issue with file name is a bit of show stopper for me however. Are you seeing the same file name truncation?
My file names tend to be long, but carry out the process for that image. Just my shorthand I guess.
thanks
Paul C
-
The issue with file name is a bit of show stopper for me however. Are you seeing the same file name truncation?
Doesn't this product ask what working space you want after processing the data? That would be a show stopper for me if this is the product I'm thinking about.
-
It asks you if you want to keep the current working space, as one of the options (states that is a faster way to operate). Or you can pick the output space.
So far I have just told it to keep current color space.
Paul C
-
I compared it at 200% with the tool posted on ll in August 2014. Image contained a dark brown negro model and an albino negro model.
So far the ai is slightly sharper in direct comparison. But also nog that impressed with how it looks. Skin looks a bit artificial on the high key.
Need to do more tests with different images.
-
It asks you if you want to keep the current working space, as one of the options (states that is a faster way to operate). Or you can pick the output space.
So far I have just told it to keep current color space.
Paul C
Thanks. But why ask? Is the process not taking place on the current color space, converting otherwise?
-
Time to process is very long, around 20 minutes to uprez a D850 image to 30 x 40 at 300 dpi or another run just uprezing to 150%. On my Macpro, 1 year old, 4 way, 64GB of ram High Sierra.
Comparing at 100% to same file run on Adobe CC "enlargement and preserving details", images appear very similar, thus for me hard to justify the 20 minute run time.
Hi Paul,
'Topaz A.I. Gigapixel' can run batches in the background while you do something else, although it uses a lot of 'horsepower' from your computer. Improvements will be added to the application, but it remains a very processor and GPU intensive procedure.
Here's an inside story about its development:
https://topazlabs.com/a-i-gigapixel-story/
Using it for modest amounts of upsampling will only make a small difference in printed output quality. But the larger the output magnification is, e.g. to print at the printer's native resolution (600 PPI or 720 PPI), the more noticeable/visible the increased quality becomes. Obviously image content and quality also play a role. Lot's of intricate detail benefits more than large uniform areas and smooth gradients, and the higher quality the input is (e.g. properly capture sharpened and low noise), the higher the upsampled quality will be.
Topaz also seems to have no way to keep the original file name, (at least I can't figure it out) as the file name is truncated to about 25 characters. You have an option for a "output" suffix, so maybe you could rewrite the full file name, but that's not what I would want to do.
This seems to be something that can be easily adjusted, just let them know. I've previously had Service requests that have been adequately resolved, so I'd expect them to tackle this as well when someone points it out to them.
Tool may have better results on jpgs or smaller files, have yet to try anything besides D850 images.
I've also tried it on large high quality images, and the result at 600% was magnificent (although the file size may exceed the TIFF limitations when 16-bit/channel quality is selected), when compared to Benvista's Photozoom Pro or On1's Perfect Resize.
Another thing one could try (e.g. with mediocre input images), is to significantly upsample, and then downsample to the required dimensions. The upsampling will add high resolution detail at the new size, which will be used to downsample to a potentially higher quality.
Another possibility is to blend 2 images in e.g. Photoshop, one A.I. Gigapixel layer on top of a regular upsampling layer, in case that produces a more pleasing result (which remains to be seen).
Cheers,
Bart
-
Thanks. But why ask? Is the process not taking place on the current color space, converting otherwise?
Hi Andrew,
We don't know the inner workings but, in general, resampling at linear gamma has benefits for color fidelity, smoothness of gradients, and can reduce halo artifacts. So sooner or later, there are bound to be interim gamma changes, and a final mapping of the result to the required colorspace.
Cheers,
Bart
-
Hi Andrew,
We don't know the inner workings but, in general, resampling at linear gamma has benefits for color fidelity, smoothness of gradients, and can reduce halo artifacts. So sooner or later, there are bound to be interim gamma changes, and a final mapping of the result to the required colorspace.
Cheers,
Bart
Good point! I know ACR/LR do the same but honors a rendered images color space. I don't think PS does however. I was wondering if this product might be converting to Lab like the old recommendations to do so in the past but you're right, we'll never really know.
-
Good point! I know ACR/LR do the same but honors a rendered images color space. I don't think PS does however. I was wondering if this product might be converting to Lab like the old recommendations to do so in the past but you're right, we'll never really know.
Photoshop definitely does not resample using gamma=1 and you can get some problematic aliasing (moire) if you have close, regular patterns, like striped shirts/picket fences in the image. When that occurs I switch to gamma=1 then resample and switch back.
-
Strange thing that I think is occurring, and experimental users please confirm this - when you pass a tagged,non-sRGB image to AIGP and use the "Match Input (faster)" color profile option, what appears to happen is that the file remains in the input color space but gets tagged with an sRGB tag upon output. When you open the file in PS, the document color space is indicated as sRGB but the color is, predictably, undersaturated. Assigning the original profile corrects the problem.
In other words, the current version as a little hiccup in terms of tagging the output.
kirk
-
I have also noticed that AIGP is doing noise reduction and some slight sharpening, presumably using the AI Clear algorithms in conjunction with the scaling of the AIGP network.
kirk
-
Photoshop definitely does not resample using gamma=1 and you can get some problematic aliasing (moire) if you have close, regular patterns, like striped shirts/picket fences in the image. When that occurs I switch to gamma=1 then resample and switch back.
Could you share how you do it? Do you have a icc profile with gamma=1 or change mode to 32 bits?
Thanks
-
I have tried it and it does a good job from what I have seen so far. If you are pressed for time and want a quick workflow this isn't for you working on large original files....it is very slow. I have other ways to enlarge that are much faster and I would say as good quality wise.
-
Could you share how you do it? Do you have a icc profile with gamma=1 or change mode to 32 bits?
You could create such a profile using just Photoshop if so desired. Select the RGB working space you want in Color Settings. Select Custom RGB... from the same menu, alter the Gamma, close and save the profile. Now SHOULD you do this? ;)
-
You could create such a profile using just Photoshop if so desired. Select the RGB working space you want in Color Settings. Select Custom RGB... from the same menu, alter the Gamma, close and save the profile. Now SHOULD you do this? ;)
Yup, that works. But be sure to use 16 bit RGB first. 8 bits will royally hose the image in linear gamma.
I've not seen the problem often. Most frequent occurrence is when printing a high rez image from modern DLSRs with a printer when the DPI doesn't match and the printer driver downsamples or when the image is downsized to the printer's native DPI. Then, downsampling to printer native DPI in linear gamma fixes the printer aliasing. It's really not often a problem which also means when it comes up people have a hard time figuring the cause and confuse the moire in the print with intrinsic capture moire which is more common. Especially with cameras that don't have anti-aliasing filters. Once it's in the image it's a pain to remove or mask.
-
Thanks Andrew, Doug
I was wondering if that was the recommended way.
-
BTW, linear gamma is best for fixing moire that is induced by downsampling. Upsampling routines don't have the problem and the better ones, like the topic of this thread, shouldn't change gamma. There are some fairly sophisticated upsampling algorithms that do things like look for edges and enhance them and they work best in normal gamma colorspaces. They are already aware, and compensate for, gamma. Linear gamma is best for linear transform operations which are typical of downsampling where the higher spatial freqs are cut and the algorithms are not gamma aware such as Photoshop's.
-
Here my first experience with Topaz AI Gigapixel:
My Mac: Mac Pro Early 2009, 2.66 GHz Quad-Core Intel Xeon, 24 GB RAM, Graphic Card NVIDIA GeForce GT 120, 512 MB. Mac OS 10.11.6.
AI Gigapixel version 1.0.1
The image I tried to test was a Canon 5D2 file ( ca 20 MP) with a little cropping.
Under Enlarge Options, ppi, I can enter a max of 300 ppi, not 360. So this would require additional re-scaling later, since I want 360ppi.
During the processing, the image size in pixels is displayed by the application as original 3454x5611, after upscaling (to 126 cm in height) 23 268 x 37 800.
After upscaling in Iridient, where I can enter the intended size in mm, the final size ist 10 993 x 17 868 px, as displayed by the Mac Finders Get Info. --??
During the processing, all sorts of weird things happen in the Finder, when I tried to move the application window or minimize it or do anything else at all. Amongst others, the Dock did not pop up any more. Every movement of the app window was extremely unsmooth, if at all possible. It seemed extremely delayed so that it was all unpredictable. Small dots (1 and 2 pixels?) appeared in the app window, in the Finder window and in the menu bar. (Fortunately, they disappeared after restart of the Mac.) During one period, the mouse pointer was grossly enlarged. The Time Machine menu was displayed (something I had not asked for), and it took a while and some other Finder clicks, not sure which, before it disappeared again.
After about 18 minutes, I decided to quit. I had to force quit, which was possible with some hassle, I think due to the unpredictable delay of actions after mouse click.
So the Mac version needs some de-bugging before I can test it any further.
-
My Mac: Mac Pro Early 2009, 2.66 GHz Quad-Core Intel Xeon, 24 GB RAM, Graphic Card NVIDIA GeForce GT 120, 512 MB. Mac OS 10.11.6.
from Topaz ...
All NVIDIA cards with <1GB of dedicated VRAM are unsupported
Might explain some of your problems.
-
on screen stuff hard to really tell when it comes to output. This file is from an IQ180, and has been uprezzed to produce a 74x37 inch print at 360 dpi. I have printed this image up to 84” directly from LR with excellent results (using Lightroom standard output sharpening). Upper left is output from LR print module to jpg, sharpening set to standard/glossy. Top right is using Topaz AI Gigapixel, bottom right is Photoshop preserve details 2 with 30% noise reduction followed by Focus Magic at 7. Bottom right is PS Preserve details 2 no noise reduction or sharpening. I added a small version of the full file to help with context of what the image is.
Took 20 minutes with 2018 MacBook Pro 2.9ghz/32gigs ram and eGPU Radeon RX Vega 56 8176 MB.
I’m going to print a strip from all 4 files to see if there is much visible difference. Not sure there will be.
-
Oops! thanks for your alert, Wayne. I would have liked to compare AI to Iridient, which I find superior in upscaling and sharpening to Photozoom Pro, which otherwise has received much praise by Bart. But before I buy into a new graphics card...
-
Oops! thanks for your alert, Wayne. I would have liked to compare AI to Iridient, which I find superior in upscaling and sharpening to Photozoom Pro, which otherwise has received much praise by Bart. But before I buy into a new graphics card...
Hi Hening,
I feel your pain, my graphics card is also not supported (although it has plenty of memory). Sometimes it works, but most of the time it crashes/closes without producing a result. Given their compatibility information (here (https://help.topazlabs.com/hc/en-us/articles/360012811791-A-I-Gigapixel-Minimum-Requirements-and-Recommended-Hardware)), it is unlikely they will add support for my card at a later date.
So, a hardware upgrade has gotten a higher priority (amongst other priorities).
Cheers,
Bart
-
Hi Bart,
if you upgrade your hardware before I do (which sounds likely ;-) ), please consider comparing to Iridient, which is now Windows-too :-)
Good light!
-
I also used an 80mp file that I have and resized it to 74 inches with A.I. With my recent Windows build the process took 30 minutes (Wayne advantage the eGPU). Topaz utilizing the GPU for a lot of work would require me to upgrade my GPU to meet or beat Wayne's times. The result is more what I have seen so far. When compared to Adobe 'Preserve Details 2.0' it is initially, maybe, a little sharper because Topaz uses some sharpening. Image darkening is an issue which has been previously mentioned. But I think that the Adobe approach is much more preferred for me because of the overall final appearance of the file. The final output is little smoother and just as sharp (with a little tweak) with the same amount of detail preservation without an artificial look - and takes significantly less time.
Adobe deserves a lot of credit for their efforts in this direction.
Victor
-
One thing is for sure, I had totally missed the "preserve Details 2" uprez in CC 2018. Thanks to Victor and Wayne for bringing it up.
No doubt that it's much better than previous versions and IMO is for what I do noticeably better than the Topaz output. Topaz may win in a 600 % uprez, but I just don't need that much as I am starting with 42 to 50MP.
Paul C
-
An update is available:
Update V1.0.2 has the following changes:
New Features:
- none
Bug Fixes:
- Custom ratios did not work correctly
- Start button was not enabled when size was measured in inches
- Installer would stall after user accepted terms and conditions
- Saved as 72ppi instead of original resolution
- Underexposure of processed images has been improved
GUI Modifications:
- none
-
The update has dramatically improved the image darkening issue. Speed also seems to have improved. I took a GFX file and upsized it to 40 inches at 600dpi which took 20 minutes. Compared to Adobe Preserve Details 2.0 I would say that Topaz has the edge. It's not a sharpening issue any longer but an overall look. Topaz has a smoother appearance. Upsizing to 600dpi would eliminate the Canon software upsizing which is incorporated into the Canon Plugin.
Victor
-
Thanks for your tests Victor. Surely we can expect Gigapixel to get better as time goes on; most software tends to.
-
I did some more tests today (http://plugsandpixels.com/blog/topaz-a-i-gigapixel-vs-ps-on1-more-tests/), this time with DSLR-sourced images and including ON1 Photo RAW in the mix.
-
I did some more tests today (http://plugsandpixels.com/blog/topaz-a-i-gigapixel-vs-ps-on1-more-tests/), this time with DSLR-sourced images and including ON1 Photo RAW in the mix.
Hi,
Very interesting results, and assuming that the rightmost image is AIG, it demonstrates that the A.I. is adding 'plausible' detail where there is none to speak of. That's the main difference between traditional interpolation between existing pixels, and using a Neural network to replace pixels.
Cheers,
Bart
-
Thanks Bart, yes, Gigapixel is the right-most of each example.
Except for the cases where images are going to be presented for legal evidence and should be left "as-is" as much as possible, I can't see any reason not to be really excited about this new upsizing tech. I am really shocked at how much better the results seem to be when compared to traditional upsizing methods.
I would like to see some results from other LL members with better original photos and equipment than I have, and with other software I haven't tested yet.
-
Although looking at a file at 100% pixels tells something there is nothing like a print to really see how this software works. I own both a GFX and a Phase 3100 and shot the same scene (front of my house that has tons of tiny plant detail) using same lens and adjusting distance to subject to account for focal length change. The GFX file was upsampled to 40 inches on the wide side at 600dpi. The 3100 file was also upsampled very slightly with preserve details 2.0 to 40 inches on the long side at 300dpi. I printed to my ipf8400 using the print plugin at 600dpi. That means that the GFX file never saw the Canon upsampling being a native 600dpi file. The Phase 3100 was upsampled to 600dpi using the print plugin - which has been my print workflow.
The final product is really kind of staggering. I'm really fussy about print detail which is the reason I limit upsampling. I would never upsample a GFX file beyond 40 inches unless the content could visually sustain the upsampling. Landscape detail already suffers at 40 inches - upon close inspection. My benchmark is my Phase 3100 which can easily handle 40 inches. The test crops were printed on Epson semi gloss photo paper which will easily show file anomalies.
To my surprise the GFX AI Giga upsampled file easily, visually, matched the detail of the Phase 3100 file. This is with very close inspection when viewing at my EVS Executive Viewing Station. Fine plant detail appears very natural and very sharp. There is an overall smoothness to the print that is very pleasing. You can't see any of this on a monitor - you have to print to see how this software works. Again........ I'm stunned! I could care less about the 20 minute wait as the results are worth it.
To think that this technology is only going to get better is very exciting.
Victor
-
Thanks Victor for your report! Can we come over and see your prints? ;-)
-
Any time you would like..... ;)
Victor
-
Hi,
No great surprise but still nice to hear about it. I got the impression that GFX has nice lenses. But, we need also discussing things like aperture, once we are past f/8, diffraction may be a limiting factor.
The GFX will need less stop down than larger formats like 54x41 mm, to keep same DoF.
Best regards
Erik
Although looking at a file at 100% pixels tells something there is nothing like a print to really see how this software works. I own both a GFX and a Phase 3100 and shot the same scene (front of my house that has tons of tiny plant detail) using same lens and adjusting distance to subject to account for focal length change. The GFX file was upsampled to 40 inches on the wide side at 600dpi. The 3100 file was also upsampled very slightly with preserve details 2.0 to 40 inches on the long side at 300dpi. I printed to my ipf8400 using the print plugin at 600dpi. That means that the GFX file never saw the Canon upsampling being a native 600dpi file. The Phase 3100 was upsampled to 600dpi using the print plugin - which has been my print workflow.
The final product is really kind of staggering. I'm really fussy about print detail which is the reason I limit upsampling. I would never upsample a GFX file beyond 40 inches unless the content could visually sustain the upsampling. Landscape detail already suffers at 40 inches - upon close inspection. My benchmark is my Phase 3100 which can easily handle 40 inches. The test crops were printed on Epson semi gloss photo paper which will easily show file anomalies.
To my surprise the GFX AI Giga upsampled file easily visually matched the detail of the Phase 3100 file. This is with very close inspection when viewing at my EVS Executive Viewing Station. Fine plant detail appears very natural and very sharp. There is an overall smoothness to the print that is very pleasing. You can't see any of this on a monitor - you have to print to see how this software works. Again........ I'm stunned! I could care less about the 20 minute wait as the results are worth it.
To think that this technology is only going to get better is very exciting.
Victor
-
I received an email today asking why I hadn't also tested BenVista PhotoZoom Pro 7 against Topaz Gigapixel. Well, here you go ;-)
-
Hi,
No great surprise but still nice to hear about it. I got the impression that GFX has nice lenses. But, we need also discussing things like aperture, once we are past f/8, diffraction may be a limiting factor.
The GFX will need less stop down than larger formats like 54x41 mm, to keep same DoF.
Best regards
Erik
Hi Erik...... I think you are missing the point. My test was shot with a 72mm Digitar - not a Fuji Lens. This test was not about lenses but about upsampling software. My Cambo is set up to accept the GFX and/or a DB. That is how I am able to test both sensors with the same lens at the same aperture. What is surprising to me is that, since I print large, I have been extremely concerned about upsampling and the results from the AI Giga have surpassed anything I have ever used. The 3100 eliminated a lot of upsampling issues but even so I am so finicky that I don't even print 48 inches long without stitching two shifted files from my 3100. To have a single 50mp file from the GFX equal a single 100mp file from my 3100 in print at 40 inches is no small feat. I've used every upsampling software there is from ON1 perfect resize to Super Zoom to Adobe 'Preserve Details 2.0 and the print is the final judge and jury. So far AI Giga is the real deal and the best there is..... only to get better.
Victor
-
Hi Victor,
How are you picking a dpi for output in Topaz? I see the scale options, 200, 400 600 percent, but way to set a dpi?
thanks
Paul
-
Use inches not pixels for the width and other options will become available which includes 'dpi'.
Victor
-
I have run across one bug that I haven't been able to rectify. If I use pixels as the width or height value then I cannot enter any more than 4 significant digest into either field. The only way so far around this is to use inches in the 'fit to' value field. I have sent Topaz a help request so will see if its universal or just me. Has anyone else experienced this?
Victor
-
thanks Victor.
Just finished a test from a D850 image. MacPro, 2 years old, Xeon 4 way, can't remember speed, 64GB of ram, and supposedly 2 radeon graphics cards.
Tried a 600%, and it took close to an hour to run, and when it finished the image was 4.9 GB, but CC could not "parse" the file and thus would not open it, never have seen such an error before in photoshop. I wonder if it had to do with the fact the file was a tif and over 4GB? will Adobe not open a tif over 4GB, I know it won't save one that large.
Ran a 2nd one at 400%, this one came out to be a 73" x 110" image at 300 dpi, (where I tend to keep everything), I do print everything from Lightroom at either 240 or 360 dpi. Did the same thing with Adobe CC preserve details 2.
Overall just looking at 100% the Topaz file is better, Adobe shows a lot more strange noise and or mottling to the image, which is very clean at base output resolution of 300 in LR. I don't understand why Topaz saves the file considerably smaller, 2.9GB and CC was over 4 GB and needed to be a PSB.
I will try to post some samples a bit later, but it's always hard to get the point across IMO.
Main issue for me now is time. It still took the same MacPro about 45 minutes to run the uprez. Not sure if some of this is due to the machine power settings, as it will always turn off the screen if left running for that length of time, and may be timing out also. Need to check my power off/sleep settings.
Will work up a few more test shots in the next few days. Biggest issue for me is resources and processing time. As Bart mentioned, Topaz is taking a huge amount of the system to run as even web browsing on that Macpro is much slower while the tool is running in the background.
Paul C
-
Thanks Paul, your tests remind me that a colleague at work has an iMac Pro which might be fun to test with Gigapixel. I also have access to 2013 Mac Pro's.
I did more onscreen tests today on the MBP, changing pixel mush into something usable.
600% upsampling of the lower left of this storefront image (shown here in reduced size for reference) as viewed in Photoshop at 100% (native pixels/Gigapixel).
-
Paul.... for some reason I get an error message from PS when opening the Giga file that says that the metadata is unreadable but the file still opens. This is even after I tell Giga to include metadata. I don't know if that is your issue.
The reason I upsample to 600dpi is because I want to bypass any and all other driver/software upsampling. I have always found this to yield the best results. If I were printing to an Epson then I would upsample to 720dpi.
Today I took an old 1ds file taken back in 2003 and processed it in PS and upsampled in Giga to 30 inches at 600dpi. I haven't printed it yet but the detail on screen at 33% pixels (which is a good screen visualization as to what the print will look like upon close inspection) is startling. I would have never upsampled that file beyond 20 inches on the long side in the past.
I do hope that Topaz fixes the bug I mentioned. I want to be able to enter pixel dimensions as I print to dimensions that are just a smidgeon more or less than standard (instead of 4X3 it may be 4.11X2.97, etc). This is due to the fact that I museum wrap and stretcher bars are never perfectly sized. I want the face of the frame to contain the image and the sides (depth of the stretcher bars) to be white.... its very difficult to accurately do and its rare that I get it perfect. I can get around all of this by using inches and printing slightly on the large side and then cropping but those are extra steps.
Please post on your observations of an actual print.
Victor
-
thanks Victor.
Just finished a test from a D850 image. MacPro, 2 years old, Xeon 4 way, can't remember speed, 64GB of ram, and supposedly 2 radeon graphics cards.
Tried a 600%, and it took close to an hour to run, and when it finished the image was 4.9 GB, but CC could not "parse" the file and thus would not open it, never have seen such an error before in photoshop. I wonder if it had to do with the fact the file was a tif and over 4GB? will Adobe not open a tif over 4GB, I know it won't save one that large.
Hi Paul,
TIFF has a limit of 4GB, anything larger should be saved to either the BigTIFF or the PSB format.
I expect that AIG will get a future enhancement to save in such formats. I also expect that they may add (if users request it), a possibility to add noise or dithering at the final output dimensions. That would prevent some of the (too) smooth featureless regions (which risk posterization in 8-bit/channel output) at maximum scaling settings.
To reduce the risk of posterization as much as possible, one could tag the source image with the output profile before scaling which avoids an additional colorspace conversion with too few bits of precision, and that allows the image to be saved as an 8-bit/channel TIFF because it is the final data with no further editing expected.
Since upgrades are free for life, I expect it will only take some time before users are rewarded with these refinements.
Ran a 2nd one at 400%, this one came out to be a 73" x 110" image at 300 dpi, (where I tend to keep everything), I do print everything from Lightroom at either 240 or 360 dpi. Did the same thing with Adobe CC preserve details 2.
I'd try 720 PPI (if the file size can be kept under 4GB), and once the preset is added (otherwise calculate the number of pixels required). This will produce even more accurate detail, and AIG will use every pixel it can to add detail.
Overall just looking at 100% the Topaz file is better, Adobe shows a lot more strange noise and or mottling to the image, which is very clean at base output resolution of 300 in LR. I don't understand why Topaz saves the file considerably smaller, 2.9GB and CC was over 4 GB and needed to be a PSB.
Don't know why, but possibly the smoother image can be losslessly compressed further.
Cheers,
Bart
-
Very interesting discussion. Thank you for the tests and the commentary. I have a few questions that seem relevant to the issue of file size and "best practices" with respect to using AIG, given the issues about the speed of the program.
I print from an Epson 9900 printer and have upressed my files, which start at 360 ppi, to the desired size, leaving the file in 16 bit format and the "native" color space (Prophoto). I just feed the 16 bit file to the Photoshop print dialog box and pick the appropriate profile for the desired paper.
Should you upress to the desired print size with AIG at 720 ppi? Should you convert the 16 bit file to 8 bit before upressing? Should you convert the file to the desire paper profile before upressing?
-
Thanks Bart, I will try the 720 ppi but that well may again take the file past 4GB on a 400% uprez.
The single biggest issue for is time. It took 1 hour to uprez a Fuji image (2 part pano) and about 45 to 40 minutes to uprez the D850 images.
IMO a bit too long.
Paul C
-
Very interesting discussion. Thank you for the tests and the commentary. I have a few questions that seem relevant to the issue of file size and "best practices" with respect to using AIG, given the issues about the speed of the program.
I print from an Epson 9900 printer and have upressed my files, which start at 360 ppi, to the desired size, leaving the file in 16 bit format and the "native" color space (Prophoto). I just feed the 16 bit file to the Photoshop print dialog box and pick the appropriate profile for the desired paper.
Should you upress to the desired print size with AIG at 720 ppi? Should you convert the 16 bit file to 8 bit before upressing? Should you convert the file to the desire paper profile before upressing?
Hi,
If the upsampled/scaled image has yet to undergo a color space conversion, e.g. from ProPhoto RGB to the output profile, I'd keep it in 16-bit/channel mode. After the conversion to the output profile, it can be made smaller by using 8-b/ch mode, even as a high-quality JPEG, because most printer driver pipelines are 8-b/ch.
If the image to be upsampled is already converted and tagged with the output profile/colorspace, then AIG will probably have no difficulty with scaling an 8-b/ch input file (JPG or TIFF) to an 8-b/ch output file.
If the final file size limit (4GB) is not exceeded, I'd try the highest native printer resolution, i.e. 600 or 720 PPI. Make sure to set the printer driver's options to accommodate the higher level of detail (i.e. "finest details" must be checked to on if an Epson is used, otherwise it will do a mediocre downsample to 360PPI).
How noticeable the difference will be, depends on the input image (subject & quality) and amount of upsampling. The 4x as many pixels will take longer to be calculated, so it's a trade-off which is only justified if the quality jump is visible enough.
Cheers,
Bart
-
Thanks Bart, I will try the 720 ppi but that well may again take the file past 4GB on a 400% uprez.
The single biggest issue for is time. It took 1 hour to uprez a Fuji image (2 part pano) and about 45 to 40 minutes to uprez the D850 images.
IMO a bit too long.
Hi Paul,
While the folks at Topaz Labs will try to improve the required processing time, it will remain a very processor intensive operation (CPU+FPU+GPU). Even on my not supported video card, the GPU processing will frequently hit a 100% load. So it will never be a fast procedure for large files on current hardware. Check out the processor load to see which component is the bottleneck. Only with faster hardware can we expect to shave off waiting time.
That's why AIG does batch processing and runs as a separate application, so as to not block other processes (although they will slow down). If the computer is set to not go to sleep, it can run overnight.
Cheers,
Bart
-
Hi,
If the upsampled/scaled image has yet to undergo a color space conversion, e.g. from ProPhoto RGB to the output profile, I'd keep it in 16-bit/channel mode. After the conversion to the output profile, it can be made smaller by using 8-b/ch mode, even as a high-quality JPEG, because most printer driver pipelines are 8-b/ch.
If the image to be upsampled is already converted and tagged with the output profile/colorspace, then AIG will probably have no difficulty with scaling an 8-b/ch input file (JPG or TIFF) to an 8-b/ch output file.
If the final file size limit (4GB) is not exceeded, I'd try the highest native printer resolution, i.e. 600 or 720 PPI. Make sure to set the printer driver's options to accommodate the higher level of detail (i.e. "finest details" must be checked to on if an Epson is used, otherwise it will do a mediocre downsample to 360PPI).
How noticeable the difference will be, depends on the input image (subject & quality) and amount of upsampling. The 4x as many pixels will take longer to be calculated, so it's a trade-off which is only justified if the quality jump is visible enough.
Cheers,
Bart
Thanks, Bart.
Does AIG do any sharpening of the file as part of the upressing process? Hopefully not.
-
I always thought they did a small amount of sharpening but can't confirm. As has been covered by Bart I always have in the past upsampled the image to be printed in my favorite program to desired DPI to avoid any other upsampling drivers/software on the printer side. This was before I started to use the print plugin for my Canon printer which upscales to 600DPI which I started to use exclusively at that point. In other words I would upsample only to 300DPI. Now with Giga I just let the program upsample to my desired print size (40 inches minimum) at 600DPI to avoid even the Canon software. The printed results have surpased any other workflow I have used.
This software will more than likely become a standard part of my workflow as any file that I want to print has to upsampled. That includes anything (even stiched) that comes out of my 3100. Phase doesn't have enough pixels now (minimum 24000 on the long side) nor will they anytime soon.
Victor
-
Thanks, Bart.
Does AIG do any sharpening of the file as part of the upressing process? Hopefully not.
Hi,
I don't think we can call it sharpening in the normal sense, and I do not see any signs of sharpening halo or ringing either. AIG is actually trying to restore the original sharpness that an image of that output size would have, had it been shot at that size.
That's how Neural Networks function, after being trained on a small version of the scene and on the full-size version of the same scene. It then learns how to replace interpolated pixels with the actual pixels that should be there. Given the possible variety of image content that it may be confronted with, it has to be trained on many different types of images.
This also means that it sometimes guesses wrong when dealing with other images than those it was trained on. This is what the folks at Topaz Labs are working on to improve, but I've read that it can take a week of additional training before they can see if a tweak gives better results or not.
Cheers,
Bart
-
Well, I tried a few on this Mac Pro at 720 dpi and the tool aborts about 3/4 of the way through. Not sure why, no error.
But still making progress.
What I have realized is the "enhanced" setting pulls does add a bit of smoothness, and pulls out all the noise in areas like sky. Holds up good edges and still has fine details.
Will go ahead and add this to my tools, and do hope they get a better handle on the speed.
Paul C
-
Hi Bart,
From what I have seen is that the algorithm segments detail and than renders those segments artificially.
Best regards
Erik
Hi,
I don't think we can call it sharpening in the normal sense, and I do not see any signs of sharpening halo or ringing either. AIG is actually trying to restore the original sharpness that an image of that output size would have, had it been shot at that size.
That's how Neural Networks function, after being trained on a small version of the scene and on the full-size version of the same scene. It then learns how to replace interpolated pixels with the actual pixels that should be there. Given the possible variety of image content that it may be confronted with, it has to be trained on many different types of images.
This also means that it sometimes guesses wrong when dealing with other images than those it was trained on. This is what the folks at Topaz Labs are working on to improve, but I've read that it can take a week of additional training before they can see if a tweak gives better results or not.
Cheers,
Bart
-
Tried AI Gigapixel on my laptop (Win10 64bit, 16GB RAM Geforce GTX 960M 4G) and got very slow but good results with AI G, upsampling 5184x3188 pixel files to 3X or 4X. I'm impressed enough with the quality, and when I get home I'd like to use AI G on my desktop but I am limited by budget issues. Here's the story:
I have an ancient Alienware Area 51 desktop running Win10 64B. It has an i-7 4 core CPU, is maxed out at 12GB DDR-3 RAM, and I'm using 2 Geforce GTX 460 DDR-5 video cards, 1G each. I am a heavy user of LR Classic CC and PSCC, and have been happy with the old computer: I use 6 internal hard drives (4 for images) as well as a bunch of external drives for backup. The computer also has a 1500W power supply and liquid cooling, and it would cost much more than I can afford to replace it. It runs just fine for my purposes.
The question: The video cards are only 1G, are not used in parallel, and are on the AI G Not Supported list. The video cards are 256-bit, PCI-E 2.0 x16, width:dual slot.
1) Am I right in thinking that I cannot use the newer PCI-E 3.0 cards (the computer was made in 2009)?
2) If I can find video cards that will work with the PCI-E2 specs, should I get one card with as much VRAM as possible?
3) Or is the PCI-E spec dependent on the video card rather than the computer slot, in which case I could get a PCI-E 3 card?
-
Tried AI Gigapixel on my laptop (Win10 64bit, 16GB RAM Geforce GTX 960M 4G) and got very slow but good results with AI G, upsampling 5184x3188 pixel files to 3X or 4X. I'm impressed enough with the quality, and when I get home I'd like to use AI G on my desktop but I am limited by budget issues. Here's the story:
I have an ancient Alienware Area 51 desktop running Win10 64B. It has an i-7 4 core CPU, is maxed out at 12GB DDR-3 RAM, and I'm using 2 Geforce GTX 460 DDR-5 video cards, 1G each. I am a heavy user of LR Classic CC and PSCC, and have been happy with the old computer: I use 6 internal hard drives (4 for images) as well as a bunch of external drives for backup. The computer also has a 1500W power supply and liquid cooling, and it would cost much more than I can afford to replace it. It runs just fine for my purposes.
The question: The video cards are only 1G, are not used in parallel, and are on the AI G Not Supported list. The video cards are 256-bit, PCI-E 2.0 x16, width:dual slot.
1) Am I right in thinking that I cannot use the newer PCI-E 3.0 cards (the computer was made in 2009)?
2) If I can find video cards that will work with the PCI-E2 specs, should I get one card with as much VRAM as possible?
3) Or is the PCI-E spec dependent on the video card rather than the computer slot, in which case I could get a PCI-E 3 card?
Hi,
Most motherboards and computers manufactured in 2013 or later probably support PCI Express v3.0. So your 2009 desktop's motherboard probably doesn't support PCI-E. Generally, PCI Express refers to the actual expansion slots on the motherboard that accept PCIe-based expansion cards and to the types of expansion cards themselves.
PCI-Express 3.0 has already been superseded by 4.0, and 5.0 is expected to be released in 2019.
So the cheapest short-term solution woulld be to find a compatible video card for your motherboard's version of PCI. However, glancing at the Topaz Labs requirements page, it looks like it that the compatible desktop graphics cards must be at least PCI-E 3.0 versions.
So it looks like your laptop is going to get busy.
Cheers,
Bart
-
PCI-Express 3.0 has already been superseded by 4.0, and 5.0 is expected to be released in 2019.
So the cheapest short-term solution woulld be to find a compatible video card for your motherboard's version of PCI. However, glancing at the Topaz Labs requirements page, it looks like it that the compatible desktop graphics cards must be at least PCI-E 3.0 versions.
So it looks like your laptop is going to get busy.
So I feared. Thanks for the reply. At 77 years old, I'm getting used to this pattern. Gotta win the lottery, I guess. ::)
-
So I feared. Thanks for the reply. At 77 years old, I'm getting used to this pattern. Gotta win the lottery, I guess. ::)
Should be compatible, PCIe Gen3 is supported on 3rd Generation of Intel® Core™ i5 and Core™ i7 CPUs
Brian.
-
@walter.sk:
PCI-e 3.0 cards are backwards compatible, so if your existing cards are PCI-e 2.0 you should be OK with any recent graphics card on Topaz Labs compatibility list. I'm running a Quadro P4000 (PCI-e 3.0) on a vintage 2010 PC with a 1st gen i7-980x CPU on an Asus P6X58D-E m/b (PCI-e 2.0) with 12G RAM, and it runs AI Gigapixel just fine, with fairly decent processing speeds. If you don't need the workstation and 10-bit features of the Quadro, a recent gaming card will probably be faster still, and certainly cheaper.
No crashes or hangs so far even on large input files and enlargements, just one glitch where Photoshop complained about an incorrectly terminated file error but then opened it without any further problems.
FWIW, I've tested Gigapixel on a number of images (only on-screen so far), including a photo for reproduction of an oil painting with heavy and highly-detailed impasto, and have been seriously impressed. The latter showed no visible loss or degradation of detail at 200% increase; at 400% some of the fine texture appeared a little 'plastic' but comparison with a magnified view of the original showed the structure of the detail to be similar, and I suspect it will probably print better than it looks.
On landscapes with distant foliage, grass or similar fine detail there was some tendency for this to appear a little 'wormy' at 400%+, but again close examination showed this to be generally comparable to the magnified original, so the artefact is probably mostly a function of the raw conversion of fine detail at the resolution limit of the camera and lens rather than anything introduced by Gigapixel.
My previous 'goto' for difficult enlargements has been Photozoom Pro 6.0, but Gigapixel appears considerably better so far.
Malcolm
-
@walter.sk:
PCI-e 3.0 cards are backwards compatible, so if your existing cards are PCI-e 2.0 you should be OK with any recent graphics card on Topaz Labs compatibility list. I'm running a Quadro P4000 (PCI-e 3.0) on a vintage 2010 PC with a 1st gen i7-980x CPU on an Asus P6X58D-E m/b (PCI-e 2.0) with 12G RAM, and it runs AI Gigapixel just fine, with fairly decent processing speeds. If you don't need the workstation and 10-bit features of the Quadro, a recent gaming card will probably be faster still, and certainly cheaper.
Malcolm
That sounds more hopeful. I think my CPU is a gen-1 i7-980 as well, but I can't check it for another few days. From what I've observed, AI G is really impressive.
-
@walter.sk:
PCI-e 3.0 cards are backwards compatible, so if your existing cards are PCI-e 2.0 you should be OK with any recent graphics card on Topaz Labs compatibility list. I'm running a Quadro P4000 (PCI-e 3.0) on a vintage 2010 PC with a 1st gen i7-980x CPU on an Asus P6X58D-E m/b (PCI-e 2.0) with 12G RAM, and it runs AI Gigapixel just fine, with fairly decent processing speeds. If you don't need the workstation and 10-bit features of the Quadro, a recent gaming card will probably be faster still, and certainly cheaper.
Hi Malcolm,
That's good to know (I didn't, so thanks for adding that info). I did read that a conventional PCI-card doesn't fit in a PCI Express-slot, so I assumed it to be the case the other way around as well. All I know is that the card's bus and the motherboard's slots need to match, and that PCI-E comes in different lengths, 1x - 16x.
FWIW, I've tested Gigapixel on a number of images (only on-screen so far), including a photo for reproduction of an oil painting with heavy and highly-detailed impasto, and have been seriously impressed. The latter showed no visible loss or degradation of detail at 200% increase; at 400% some of the fine texture appeared a little 'plastic' but comparison with a magnified view of the original showed the structure of the detail to be similar, and I suspect it will probably print better than it looks.
On landscapes with distant foliage, grass or similar fine detail there was some tendency for this to appear a little 'wormy' at 400%+, but again close examination showed this to be generally comparable to the magnified original, so the artefact is probably mostly a function of the raw conversion of fine detail at the resolution limit of the camera and lens rather than anything introduced by Gigapixel.
My previous 'goto' for difficult enlargements has been Photozoom Pro 6.0, but Gigapixel appears considerably better so far.
My impressions are the same, although I'm somewhat handicapped with non-compatible hardware. Photozoom Pro 6 did better than most others, until Gigapixel was released. It looks like we have a new King. The King is dead, long live the King.
Close inspection of the file on display will not accurately tell how it looks in print. Print usually looks less artificial.
Cheers,
Bart
-
Hi Bart,
PCI and PCI Express are not compatible either physically or electronically. The PCI Express slots on my Asus m/b are PCI Express 2.0 x 16, while the Quadro P4000 is PCI Express 3.0 x 16; Walter's GTX 460 cards are PCI Express 2.0 x 16, so he should be OK with any PCI-e 3.0 (or higher?) replacement that satisfies the A.I. G minimum specs or better.
I'll try a test print or two from A.I. G when I get a moment and report back. I can also run a few timings if anyone is interested in the performance on my gear; despite the older CPU and relatively limited RAM it seems to hold up pretty well compared to other reported results, with processing times generally ITRO 10-20 min or so for enlargements from D850 files. No doubt the Quadro P4000 compensates for the other deficiencies in the system; it wasn't cheap but has usefully extended the lifetime of the build for several years more at least (my other primary uses for the Quadro are Capture One processing and 10-bit Photoshop display for art repro).
I'm wary of potentially using A.I. G for repro work at this stage apart from special use cases, simply because it is said to generate 'plausible' rather than actual detail in enlargements, but I've seen nothing yet far that actively disqualifies it from consideration for that purpose. For anything else that is not so reliant on the absolute integrity of the result, it looks to have a lot going for it.
Cheers,
Malcolm
-
Hi Bart,
PCI and PCI Express are not compatible either physically or electronically. The PCI Express slots on my Asus m/b are PCI Express 2.0 x 16, while the Quadro P4000 is PCI Express 3.0 x 16; Walter's GTX 460 cards are PCI Express 2.0 x 16, so he should be OK with any PCI-e 3.0 (or higher?) replacement that satisfies the A.I. G minimum specs or better.
Ah, now I get it. So his older motherboard apparently was already a PCI Express (2.0). Then things look more promising.
I'll try a test print or two from A.I. G when I get a moment and report back. I can also run a few timings if anyone is interested in the performance on my gear; despite the older CPU and relatively limited RAM it seems to hold up pretty well compared to other reported results, with processing times generally ITRO 10-20 min or so for enlargements from D850 files. No doubt the Quadro P4000 compensates for the other deficiencies in the system; it wasn't cheap but has usefully extended the lifetime of the build for several years more at least (my other primary uses for the Quadro are Capture One processing and 10-bit Photoshop display for art repro).
Would be interesting, because many people report slow speeds, but they often don't mention with which specific hardware. Also, it might be useful to note the maximum % of the "Highest utilization across all GPU engines" that's reported in the Task Manager. Mine frequently hits the 100% ceiling, suggesting that with faster GPUs one would reduce the total time required. And maybe the Quadro architecture has also other benefits for this specific use case.
I'm wary of potentially using A.I. G for repro work at this stage apart from special use cases, simply because it is said to generate 'plausible' rather than actual detail in enlargements, but I've seen nothing yet far that actively disqualifies it from consideration for that purpose. For anything else that is not so reliant on the absolute integrity of the result, it looks to have a lot going for it.
Yes, although 'regular' interpolation also produces fake/plausible information. My expectation is that AIG on average gets closer to the 'ground truth' than the more blurry methods, especially on the type of detail that human vision relies on, i.e. edges and lines. The only way to be absolutely sure of correct detail is by shooting at a larger magnification (closer up, or with longer focal length), which may then require stitching for a wide enough Field of View.
Cheers,
Bart
-
Hi Bart,
I'm inherently a purist for repro work and I do regularly stitch for larger artworks; routinely a 3-way flat stitch with a PC-E Micro Nikkor 85mm shift lens and sometimes a multi-column version of same for the largest. PTGui Pro (now v 11) does a great job on the stitching. That and a good RIP takes care of most things.
But there are times when something like A.I. G could potentially be a very useful addition to the toolbox, especially for extreme enlargements or when clients need a print from their own historic files or transparencies where the original work can't be re-photographed to the necessary standards. In such cases any potential minor inaccuracies in reproduction, while not desirable, are usually an acceptable trade-off for getting a print at the required size and of a satisfactory quality.
Parenthetically, and not wanting to derail this thread, I've experimented with regular panoramic stitching for even higher resolution but haven't found the results satisfactory for artwork. Possibly it's a function of the lens used or I haven't nailed the right technique yet, as I know it has been used for large digitisation projects.
Cheers,
Malcolm
-
Would be interesting, because many people report slow speeds, but they often don't mention with which specific hardware. Also, it might be useful to note the maximum % of the "Highest utilization across all GPU engines" that's reported in the Task Manager. Mine frequently hits the 100% ceiling, suggesting that with faster GPUs one would reduce the total time required. And maybe the Quadro architecture has also other benefits for this specific use case.
Cheers,
Bart
Hi Bart..... I have never seen 100% GPU usage in Task Manager but do hit 75% but only for milliseconds. This is with a 1070 series card. My times are routinely 20 minutes for upsizing a 50mp image to 40 inches on the long side at 600dpi which equals 24000 pixels. The size of the original file has a lot of influence on times. I've toyed with upgrading the GPU to the new 2080/70 series but really don't know if it would be worth it.
Victor
-
Hi Bart..... I have never seen 100% GPU usage in Task Manager but do hit 75% but only for milliseconds. This is with a 1070 series card. My times are routinely 20 minutes for upsizing a 50mp image to 40 inches on the long side at 600dpi which equals 24000 pixels. The size of the original file has a lot of influence on times. I've toyed with upgrading the GPU to the new 2080/70 series but really don't know if it would be worth it.
Hi Victor,
If the GPU utilization doesn't hit the 100% ceiling, then there is some reserve left and a different GPU would probably only make a significant change if it uses much higher clock frequencies, or faster memory and bus speeds. Of course, there may also be other architecture improvements possible, but at least the GPU utilization is not determining the limit in your case.
Cheers,
Bart
-
This does appear to be interesting software. I just posted a comparison/example at http://paulroark.com/AI-Gigapixel-sample.JPG . For large files (more than a single Sony a7rii frame) I ran into insufficient RAM and other issues. (I'm on a Microsoft Surface Studio.) Also, I noted on an old photo I'd copied for someone, there was no improvement in quality -- garbage in, garbage out.
I think some of the smoothed areas in the sample I worked on were too smooth. I don't know if there is any adjustment to stop this over-smoothing, which, in effect, eliminated some image content and made the output look a little too much like plastic. I think I'll try using standard bicubic on the original, put is as a lower layer, and then seeing if revealing a little of the original helps the "over-smoothed" areas.
(Note that the Sony I used for the posted sample has the KolariVision Ultra Thin cover glass. Leica M wide angles do not work well on Sony cameras with the OEM thick cover glass.)
Paul
www.PaulRoark.com
-
This does appear to be interesting software. I just posted a comparison/example at http://paulroark.com/AI-Gigapixel-sample.JPG . For large files (more than a single Sony a7rii frame) I ran into insufficient RAM and other issues. (I'm on a Microsoft Surface Studio.) Also, I noted on an old photo I'd copied for someone, there was no improvement in quality -- garbage in, garbage out.
Hi Paul,
Absolutely true, garbage only becomes bigger. But on the other hand, quality also increases. So it pays to use good quality input, properly capture sharpened (without halos and other artifacts), and Clarity applied for effective local contrast enhancements. Noise reduction, I'd be careful about that, because there is information mixed in with the noise. And A.I. Gigapixel effectively suppresses meaningless noise in favor of meaningful detail.
I think some of the smoothed areas in the sample I worked on were too smooth.
Judging from the linked example you posted, I'd say that actual detail with noise was pretty well preserved as only detail, and noise without detail was traded for lack of detail.
I don't know if there is any adjustment to stop this over-smoothing, which, in effect, eliminated some image content and made the output look a little too much like plastic. I think I'll try using standard bicubic on the original, put is as a lower layer, and then seeing if revealing a little of the original helps the "over-smoothed" areas.
Yes, one could consider a low opacity blend of some of the 'detail' as rendered by e.g. Photoshop or similar, or adding a little amount of fine grain that can fool our eyes into seeing non-existent detail. That is currently not a built-in feature that AIG has, but who knows, they might add it in the future.
Cheers,
Bart
-
@walter.sk:
PCI-e 3.0 cards are backwards compatible, so if your existing cards are PCI-e 2.0 you should be OK with any recent graphics card on Topaz Labs compatibility list. I'm running a Quadro P4000 (PCI-e 3.0) on a vintage 2010 PC with a 1st gen i7-980x CPU on an Asus P6X58D-E m/b (PCI-e 2.0) with 12G RAM, and it runs AI Gigapixel just fine, with fairly decent processing speeds. If you don't need the workstation and 10-bit features of the Quadro, a recent gaming card will probably be faster still, and certainly cheaper.
I upgraded my 2011 i7 2600K machine with a new Geforce 1070 Ti (and beefier power supply) to test AIG. My MB is Asus P8Z68-V Pro and the video card slot is PCIEx16 v 2.0.
No crashes or hangs so far even on large input files and enlargements, just one glitch where Photoshop complained about an incorrectly terminated file error but then opened it without any further problems.
AIG runs fine but the image display in CC 2018 ACR gets corrupted. I can restore it by resizing the image size in ACR. A minor headache. PS itself doesn't have any problems.
I'm not processing large files. What I am interested in, thus far, is processing old 2002ish low megapixel images so I can print them larger than 5x7. And to look decent on a 4K display. For the latter, I decided that AIG wasn't really needed--Win 10's own resizing looks decent. AIG processed images only look marginally better. They do look better if I zoom into an image, but I'm not sure if reprocessing all my old images is worth the bother for viewing on a monitor/TV.
I haven't done any printing yet. But I am enthused about AIG and I'm sure that my old four megapixel images can now be credibly printed larger than 5x7. AI driven computational photography rocks. AIG as it currently stands is only the beginning.
[Edit: I didn't upgrade my computer to process four megapixel images. I'm using those to quickly get a feel for what AIG does. I have found that the higher the quality the image the higher the quality of the AIG enlargements. I do have a Canon 6D, several 24 megapixel APS-C bodies and several L lenses and am planning on doing some new shooting at the best of my cameras' ability. With my existing equipment and old school image processing I kind of topped out at 16x20. I'm confident that starting out with high quality images processed with AIG that I can print larger than I ever did before.]
Ah, I'm going to offer an opinion. I can't understand why owners of expensive cameras are begrudging spending money on a decent computer to handle computational photography processing. If you own a high end (digital medium format) camera maybe these are the the kinds of computers you should be considering (https://www.bhphotovideo.com/c/search?setNs=p_PRICE_2%7c1&Ns=p_PRICE_2%7c1&ci=16360&srtclk=sort&N=4110474296). Along with a 32 megapixel (8K) monitor (https://www.bhphotovideo.com/c/product/1324922-REG/dell_up3218k_ultrasharp_32_ultra.html). We living in a computational photography world now. The computer is now as important as cameras and lenses. [Edit: I am planning on buying a $5,000ish computer sometime next year for AIG style post processing. My current computer suffices for experimenting. I spent $600 upgrading to just try out AIG.]
-
I posted some new iPhone 6 upscaling tests (http://"http://plugsandpixels.com/blog/topaz-gigapixel-improving-iphone-photos/"). Done on a 2013 MacBook Air ;-)
-
A couple more 600% A.I. Gigapixel (http://"http://plugsandpixels.com/discounts.html") upsizing examples, with the results at right viewed at 100% (compared to a 600% blowup of the original pixels in Photoshop).
The beach photo is from a 1-megapixel source (Kodak DC265, taken 1999).
This 4-megapixel example of the female reporter is from a Minolta DiMAGE S404, taken January 2006.
The San Diego waterfront photo is from a 6 megapixel Sony DSC-H2, taken 2007.
The original full-frame images (reduced) are included for reference.
-
Ive been testing this new A.I software over the last week. It just might be the most significant piece of software since Photoshop. It is a game changer in several ways. And not just because it opens up a world of large print opportunities, but because of what can now be done with an image downloaded from the web. Yesterday I wrote a brief piece about it on my site Blog (http://blog.jholko.com)
-
Thanks Josh, and great work on your site BTW!
-
Some feedback on AIG v 1.1.2:
I am quite happy using an ancient quad core Q8300 running at 2.5GHz on Windows 7 with 8GB of Ram. This works well because I also use a Radeon RX 580 video card with 8GB of video RAM.
This video card might well be the cheapest way of upgrading an old system and still giving decent performance on graphic intensive work.
I use Capture One v11 with no apparent slowdown. DXO Prime is slow but it seems to be slow everywhere.
Moving on… I downloaded the trial version of AIG v 1.1.2 and used it to do a 400% upres on a 35 megapixel Tiff to a 567 megapixel jpeg. The original was a scan from MF film. I chose “reduce noise strong” and “processing with GPU”.
It took a total of 22 minutes to complete. CPU usage never exceeded 4GB and I noticed no reduction in response while doing normal web-surfing. I had no way of measuring GPU usage but I increased the GPU fan speed to 80% (whatever that means) from the normal 40% to keep GPU temperature below 70C.
All in all, a good performance. The output file was obviously much bigger than the input and it looked identical in Photoshop CS6 after adjusting for file size. I did not see any of the miraculous improvements in resolution touted in Topaz’s adverts. I am also not sure that the result would have been any different when printed from the output that I get from Qimage.
Whatever, I consider it usable on a CPU constrained system but will experiment further on its overall usefulness.
-
Thanks Ian for your feedback!
As for results comparisons between PS and Gigapixel, my own tests with the initial release showed quite a difference (http://plugsandpixels.com/blog/topaz-a-i-gigapixel-vs-ps-on1-more-tests/), at least on screen where the pixels can't hide.
-
Moving on… I downloaded the trial version of AIG v 1.1.2 and used it to do a 400% upres on a 35 megapixel Tiff to a 567 megapixel jpeg. The original was a scan from MF film. I chose “reduce noise strong” and “processing with GPU”.
It took a total of 22 minutes to complete. CPU usage never exceeded 4GB and I noticed no reduction in response while doing normal web-surfing. I had no way of measuring GPU usage but I increased the GPU fan speed to 80% (whatever that means) from the normal 40% to keep GPU temperature below 70C.
All in all, a good performance. The output file was obviously much bigger than the input and it looked identical in Photoshop CS6 after adjusting for file size. I did not see any of the miraculous improvements in resolution touted in Topaz’s adverts. I am also not sure that the result would have been any different when printed from the output that I get from Qimage.
Whatever, I consider it usable on a CPU constrained system but will experiment further on its overall usefulness.
From what others have posted, I don't think AIG's models are trained for film. Others have reported that AIG isn't an improvement compared to conventional upsampling (like QImage). The AIG models are trained on digital images.
I hope they train future models for film because I have a lot of scanned film images I'd like to upsize.
-
Wayne, this may be so, but that doesn't stop one from getting great results. See here for my own tests (http://plugsandpixels.com/blog/topaz-a-i-gigapixel-challenge-upsizing-old-film-scans/) with various format films.
-
Wayne, this may be so, but that doesn't stop one from getting great results. See here for my own tests (http://plugsandpixels.com/blog/topaz-a-i-gigapixel-challenge-upsizing-old-film-scans/) with various format films.
But your images are exceptionally high quality with a lot more real image detail than film grain! I should have limited my statement to say that AIG doesn't extract image detail from film grain. (Does it?) i.e., it is a lot less useful for grainy film images than it is for noisy digital images. (Isn't it?) It can extract image detail from digital noise better than it can extract image detail from film grain. (Can't it?) That Topaz hasn't trained their machine learning models on grainy film images (Have they?)
-
I see what you mean. At this point, film grain is enhanced along with the other details.
I haven't tried using Topaz AI Clear (or DeJPEG, etc.) first on my grainiest B&W images but that would be interesting to see.
-
Wayne, based on your questions I did some tests and posted the results here (http://plugsandpixels.com/blog/topaz-a-i-clear-a-i-gigapixel-vs-film-grain-they-win/).
Turns out using AI Clear + AI Gigapixel works great for removing grain and preparing details before upsizing!
-
Further to my post of September 21, 2018 I tested the AIG software on a digital file.
On the same computer system, I took a 12 megapixel 16 bit Tiff shot on a D300 with no extra processing and used AIG to do a 400% upres to a 195 megapixel jpeg. Again I used “reduce noise strong” and “processing with GPU”.
It took a total of 7 minutes. GPU temperatures did not increase much and I detected no adverse slowdowns with web use.
The results were simply spectacular. The output was very crisp and without the halos I have seen in other examples. It puts new life into those files.
I am now going to put my money where my mouth is and buy the product (no, I have no association with Topaz or Plugsnpixels). I do hope that Topaz will address enhancing film files as I see a huge potential for restoring old film photos.
-
Thanks Ian for your feedback!
As for film files, it appears Gigapixel already handles them well (see my blog post).
And don't miss your discount!
-
Hi,
I purchased topaz AI cleared, but I'm really disapointed by the preview with high rez file (>20Megpixels) in topaz studio.
100% preview doesn't show the real image rendering (even without rendering...), which seems very inconvenient for fine tuning.
Here is a 60MP file : open in topaz and photoshop at 100%
Maybe I'm doing something wrong ? (I don't see where, I just open a file in topaz and click on 100%). Switching to low and high rez parameters in preferences has no effect.
It affects only the preview, if I save back the file, the file is as sharp as the original.
-
Maurice, are you talking about AI Clear or AI Gigapixel? This thread is about the latter. AI Clear is part of Topaz Studio, but AI Gigapixel is not. It's a stand-alone program.
-
I'm talking about AI clear, inside topaz studio, yes.
There is no preview with AI gigapixel.
-
Maurice, please break your question out to a new thread so we can help you there.
-
Maurice, please break your question out to a new thread so we can help you there.
I didn't want to make a whole thread (and give bad publicity) for maybe just a bug on my computer.
-
No problem, that's fair. And we did mention AI Clear along the way so I guess it's relevant ;-)
So is this basically a preview problem (ie, where you're seeing something different in the preview than you do with output)?
-
Yes, the preview at 100% of High resolution files is different than the real input at 100%, as you can see it in photoshop for exemple. See the image I attached a few posts before.
Is it the case for you too ? I'm window10 64b user.
-
No, Mac user (though I can run Win10 as well), but I don't have anything that hi-res. Can you provide some sample? I can check it tomorrow.
-
Something like that is enough to see a big difference at 100% :
https://www.dpreview.com/reviews/image-comparison/download-image?s3Key=5719cb5cdfcf42139e014113cabae092.jpg
-
Maurice, I believe I see what you are talking about. Images viewed in AI Clear at 100% are much less sharp than they are in Photoshop or other programs. I believe that clicking the HD Preview button at the top used to make a difference, transitioning from a rather unsharp image to one that looks similar to the Photoshop view. But clicking that button seems to do nothing now.
I am also on a Windows 64bit machine.
-
I too, have this problem Maurice. When in Studio, my previews are always kind of fuzzy when zoomed into 100%! How can anyone notice the nuances in making minor changes in contrast or sharpness.... When taking the image back to photoshop the changes can be seen very clearly now at 100% I'm working on a 4k monitor. If anyone has some input, it would be appreciated.
-
ok, it's what I thought, it's not a bug. It's annoying for fine tuning an algorithm like AIClear.
I'm on a regular 24"screen (1920x1200). I think it's even more obvious than on a 300dpi screen.
I made a ticket on the website, I don't know if they will reply.
I can understand that algorithm can be slow on high resolution file in preview, but there is other strategies to keep control on fine tuning than the way they do in their graphic interface.
it makes AI clear rendering as a black box on HD files, meanwhile theri algo seems very powerfull (I tried topaz gigapixel too).
-
Maurice, I did a quick test with Clear and 100% HD preview. It seems the results in PS are just slightly clearer than the preview at the default settings.
Let us know what becomes of your ticket.
PS: Thanks for the test chart! Though it is a bit soft (probably a 4x5 film capture).
-
Oh, It's better than what I got. :-\
-
So you'll want a Mac then, ha!
-
To be more clear, our images look like Medium quality Jpegs in Topaz Studio. This happens to me in precision clarity, detail and so on.
-
Thanks Jim, I will make them aware of our discussion.
-
Maurice and all, I thought the problem might have popped up with the latest version (1.12.0) of Studio, so I checked it on another PC that has an older version (11.5.0) installed. I see the exact same problem there, so it's not a recent version issue; nor is it restricted to just one of my computers.
I also checked to see what happens when other, non-Studio, Topaz plugins are called up directly from Photoshop rather than going through Studio. No problems there, the 100% view looks the same as it does in PS.
The HD Preview button, when invoked in Studio, induces a marked change when the image is set to FIT the screen--it becomes much sharper. After clicking the IN button one or a few times, the effect is still there but reduced. By the time I get to 100% enlargement, the HD Preview effect is completely gone.
Since Plugsnpixels' results on a Mac look good, and those of us reporting this problem are all using PC's and Windows, maybe it is the version that runs on PC's that is at the root of it. I'm wishing now that this problem was in it's own thread, so that others might more easily notice it and weigh in with more data points.
Maurice, please do let us know what happens with your support ticket. This is a pretty serious issue.
-
I heard back from Topaz about the preview and here's what they say:
"The HD Preview item is built for systems that don't have the computing power to render 100% of the effect on an image when you're in fit view. If a computer can't render the effect without significant speed loss on a large image in fit, the program downsamples the image to be able to show the effect on the full image quickly. The HD Preview version, viewing at 100% zoom and exporting all use the same render method without downsampling the preview So if you're viewing at 100% you're not going to see a difference between HD preview and not.
"The effect of a non hd preview and the actual output can sometimes vary quite a bit on the more GPU intensive artistic effects like Impression, Smudge, A.I. Remix and Glow. In those instances, we advise users to work with HD preview enabled to get a true render of what their output will look like in the canvas if they're working on any magnification less than 100%.
"What they're seeing isn't a bug, it's the way the program is intended to work to allow for quick preview for most users on slower or underpowered machines. If they're not seeing results like they expect use the HD preview mode."
-
I expexcted à behavior like this.
I would have prefered à slow behavior at preview than an altered preview.
My machine : core I7 6700k, 24go, ssd, gtx 1050, win10 64b.
-
I agree, Maurice. There should at least be an option in Preferences to select an accurate 100% preview at the expense of slower performance.
Still not sure why Plugsnpixel's 100% preview looks so much better than ours.
-
I just did a new test for you. At left is the result; at right is A.I. Clear in Topaz Studio with before-and-after split view (all at 100%). In this case I used Medium strength settings to be different from the first test. 2013 MacBook Air and High Sierra.
I think I'm seeing little more sharpness in the final result as compared to the "after" in the split view.
At any rate, the processed version of this mushy test chart is an improvement!
-
By curiosity, what is your graphic card ?
-
Intel HD Graphics 5000 1536 MB
https://www.notebookcheck.net/Intel-HD-Graphics-5000.91978.0.html
-
I think I'm seeing little more sharpness in the final result as compared to the "after" in the split view.
At any rate, the processed version of this mushy test chart is an improvement!
Thanks for continuing to look into this. I agree with you on both counts. Your processed version is better (there hasn't really been any dispute about that; AI Clear is definitely a valuable tool), and your split screen image shows a difference between before and after at 100% preview. As I read it, Topaz's explanation doesn't seem to jive with that. They say that by design HD Preview doesn't work at 100%, but it seems to for you.
So now the question is, why don't those of us on PC's see something similar to your results? Equipment differences? My GPU is an EVGA GeForce GTX 960 with 2GB RAM. CPU is an Intel Skylake i7 6700K. 32GB of DDR4 2400 RAM. Monitor is an NEC PA272W, resolution 2560 x 1440.
-
I'll refer my Topaz contact back to this thread for further comment.
-
That would be very kind. I can imagine they won't fix it tomorrow.
-
This Topaz tool looks very promising. However after looking at the following review I was not so sure anymore:
https://youtu.be/LGWEyG4DUOM
It looks like Topaz is creating all kinds of very weird things that does not belong to the image in the first place.
What do you guys (and gals) think what's happening here?
Regards,
Jaap.
-
I'd seen this video before. Well done, but be aware, he's using version 1 from a month ago and really should do a revised version. Topaz software improves quickly and continually.
-
It's à neural network algorithm. There is some failures, it doesn't work perfectly anytime.
It can work bad with test chart, but it works better on natural scenes, especially landscape, faces, etc, due to the training.
-
Would someone use AI gigapixel to make 24 megapixel images look better/sharper/cleaner when enlarged for bigger prints?
-
Would someone use AI gigapixel to make 24 megapixel images look better/sharper/cleaner when enlarged for bigger prints?
Hi Jim,
Yes, I would.
Cheers,
Bart
-
Yes, Gigapixel is designed with a dual purpose (helping salvage poor smaller images and making monster enlargements out of already hi-res ones).
-
After reading quite a lot of reviews on A.I. GP yesterday, including this thread and watching the YouTube video referenced a few posts up, I just finished demoing the product and ended up buying it. The YouTube video was a bit frightening to me, honestly, but my fears have been abated. I suspect some artifacts generated in that reviewer’s downsized resampled test image were to blame for the poor results, that and/or improvements made to the software during the last few months. I didn’t yet experience that kind of distortion.
I’ve used quite a few upscaling routines over the years, and happy to have this in the toolbox. I’ll also be looking forward to seeing it improved over the months and years as Topaz does with its products. I have an 8 core processor operating at 4.5 GHz and a mid/high range GPU that is supported and was not bothered with the processing times - generally a number of minutes for the more challenging jobs.
I do have a few preliminary observations and questions on when to use it and where it best fits in the workflow.
1. I hope it can be integrated soon into Photoshop.
2. It made me skim through my last 15 years of digital files wondering what would now print big well. I will doubtlessly spend more time doing that.
3. For 200% or less enlargement, maybe up to 400%, it seems to me there is little if any practical difference between using this product and Photoshop’s new preserve details algorithm or other premium enlargement programs. The Photoshop preserve details function I have found excellent, and I’ll continue to use that for small scaling jobs, at least until A.I. GP is integrated well into Photoshop and I become convinced it’s better there.
4. I wish there was some control over how aggressive the A.I. enhancement is. For example, in the hair and a beard of a man in an old 5MP photo I scaled up at least 10 times, A.I. GP recognized most of the hair as hair and rendered it well, but left large parts of the finer detail blurry. That can be fixed manually with a clone brush, so the picture I was experimenting with was workable, but I can imagine a slider might save quite a bit of work.
5. There are always trade offs on where to put things in the work flow. It seems this product may need to go very early on, earlier than others because it is replacing pixels instead of averaging them. I think it may go best after doing lens and abberation corrections, WB, and stretching out the histogram properly in a raw converter. I haven’t experimented where noise reduction fits in. I’m imagining some noise reduction at low levels might work best, with most everything else after enlargement, probably curves, contrast, clarity and color after enlargement as well. Especially sharpening, deconvolution or otherwise. Interested in other’s thoughts.
6. I noticed — I thought — superior performance on raw files as opposed to JPGs or TIFFs. Is that right? I’d rather not do that at least for white balance and histogram management reasons at least, and operate the software on TIFFs. Again, interested in other’s experience here.
-
Thanks Brad for your detailed thoughts.
That video (which was very well done) ideally should be removed or at least updated to more accurately reflect the state of Gigapixel at the present time. That's the thing with software, it's (usually) always improving and an outdated trail of reviews isn't helpful for prospective customers.
I'm a bit guilty of that on my own (main) website but try to keep current in my latest blog posts (a format that lends itself to staying more immediately relevant). It's all a lot of work and hard to keep up with an always-evolving type of technology!
As for direct integration with PS, Topaz has long since moved away from the plug-in model (as have most developers of that type of software) in favor of their own standalone solutions, incorporating more and more of their older titles into Topaz Studio adjustments. I think that's smart, not tying your fortunes so directly to another company and their whims.
Of course you can bring your Topaz results in and out of PS as desired.