Gigapixel v5.3.2

Topaz Gigapixel AI Releases
102

I tried using the Art mode for some old Calvin & Hobbes comic strips. The results look good for most part but the model seems to clean the intended colorful (sometimes B&W also) dots in the original image e.g. the yellow dots in Calvin’s hair or the black carbon particles around the smoke from the spaceship. These details are somewhat retained in the Architectural and Standard model but the rest of it is not as good as Art.

Here is a screenshot of the original:

i_view64_PnBDNAp0gw
i_view64_PnBDNAp0gw459×907 522 KB

and this is the 4x scaled version using Art (happens at other scales too):

Art%204x
Art%204x569×1109 982 KB

For comparison, I am adding the Standard version here too:

i_view64_R55C14SY0F
i_view64_R55C14SY0F459×896 703 KB

103

What happens with Suppress Noise set at 0, and Remove Blur set at 100?

104

Split the original image into 4 quadrants (or more if necessary), upscale them, then merge them back together.

105

Bug Report:

PS Plugin does not work, image will returned in same size.

106

Not on my Windows 10 PC - the image is returned at the expected size.

1 Like
107

My Ps 2021 plugin vers on Win 10 upscales okay too

1 Like
108

:thinking: i will recheck tomorrow.

109

Can you send me your logs?

110

This time it worked.

But here are the logs.

2020-11-30-17-34-43-Main.txt (757.4 KB) 2020-11-30-17-34-43-QML.txt (164.6 KB) 2020-11-30-17-43-31-Main.txt (44.2 KB)

111

Bug: Does not Save Model Name to Filename.

112

Confirmed.

113

Did send a 15 mb logfile (app hang zip), did force close after compute 217 tiff images.

Link was this: (https://www.dropbox.com/request/HtP7u0dAJ9zIoBmHJt9J)

114

If you mean is it adding a suffix with scaling, then it’s working perfectly on a MacBook running Big Sur

115

Performance Number for a Quadro RTX 5000 (ProGPU) and RTX 3080 (Gaming GPU).

Both 7 - 8 sec for 1080p to 4K.

If in the Future the Tensor Cores get used the numbers will not change much here (Quadro RTX 5000 to RTX 3080. (but maybe with RTX A5000 vs RTX 3080) because:

Peak FP16 Tensor TFLOPS with FP16 Accumulate.
RTX 5000: 89 Tflop
RTX2080: 84 Tflop
RTX3080: 119/238 ← second is Sparse Feature.

Peak FP16 Tensor TFLOPS with FP32 Accumulate.
RTX 5000: 89 Tflop
RTX2080: 40 Tflop
RTX 3080: 59.5/119 ← second is Sparse Feature.

Just to feel where things really work:
You know there is also the Nvidia A100 its Data Center GPU, look at the numbers here.

Peak FP32: 19.5 TFLOPS
Peak FP16: 78 TFLOPS
Peak TF32 Tensor Core: 156 TFLOPS | 312 TFLOPS2 ← second is Sparse Feature
Peak FP16 Tensor Core: 312 TFLOPS | 624 TFLOPS ← second is Sparse Feature

Price: 9.125 €

116

Has Any Body Tried Scaling 6 Saving PNG With Quality 3 Gotten a File size smaller than 2 gig Mine are 2 Gig and above.

117

I scaled one photo to 3936 x 2964 px (6X) and saved as a png with a quality setting of 3 and it is 16.9MB file size.

118

For awhile, the default “other scaling” was changed to 3X. This was great and I used it a lot. It now seems to be back to 2X. Why have a 2X default when there is already a 2X button?

119

Bugreport: Standalone and PS-Plugin does not load “correct” Auto values, you have to make auto on and off to get the “correct” values.

Win 10.

120

Giga Pixel Now when Producing File from Raw Camera File Scale 6 Setting 3 PNG now making File Greater 8 bit tif Smaller than Tiff 16.Size is 3 gig Plus Both Windows and Mac

121

Ha! got it.

Again, Gigapixel did return original size with detail enhancement and not the 20.000 i typed in.

2020-12-04-16-27-57-Main.txt (56.1 KB) 2020-12-04-16-27-57-QML.txt (21.5 KB)

Update: i did not type in, i did use what i did type in before (Software start before (yesterday) and just pressed ok, i’ve to press enter into the width field to get it loaded the resolutions.