Topaz Sharpen AI, what changed in v 1.3?

[earlybird]

Hi,
 I have been using Topaz Sharpen AI a bit and the latest 1.3+ update seems to have added an auto analysis and perhaps modified the orientation of the sliders.

 I do not think the function has changed, but I am trying to recall or understand how apparent or imagined changes with the parameter slider positions relate to previous versions. When I installed the new version the previous version was no longer available to view so I can not make a comparison.

 I am wondering if someone has some insights about what changed and can comment on the subject.

 Thank you!

[plugsnpixels]

v 1.3.0
July 17, 2019

New Feature:
Auto Parameter estimation

Change:
Improved models, can handle larger motion blur and focus blur

[JaapD]

For me in real practice: less artifacts. This release just might be good enough to actually apply to my images and do better than FocusMagic. It’s about time since I’ve purchased the complete AI package and so far I’ve found it not yet usable.

Regards,
Jaap.

[Bart_van_der_Wolf]

For me in real practice: less artifacts. This release just might be good enough to actually apply to my images and do better than FocusMagic.

I agree. The artifacts from the prior AI models didn't please me, but the newer models exhibit less stair-stepping of high contrast edges/lines. The auto parameter setting is usually close, but might still require a little tweaking. But that tweaking is much faster when already in the ballpark.

The ability to remove subject motion or camera shake is much more powerful than what can be done with Focusmagic. I'm still hoping for an upgrade (version 2.x?) that incorporates masking, so the sharpening can be targeted to specific regions of interest.

Cheers,
Bart

[earlybird]

Hi,
 From what I recalled the previous version had a scale that was negative - zero - positive, with zero at the midpoint of the slider, but now it seems like it is zero - fifty - one hundred, with fifty at the midpoint.

 I just went and looked at some older screen shots and see that I was mistaken. It is just the automated positioning of the slider that is making me think the scale has changed.

 The previous default position at the midpoint "50" is now often times automatically set to a value below "50". I had come to think of "50" as the "zero" point, with "more" or "less" being an adjustment I made. I guess the habit was so ingrained that I regarded the midpoint as "zero" without thinking much about it.

 Sharpen AI works great, but my engagement with the interface felt different. A screen shot of a previous version was all I needed to answer my questions.

 Thank you.

 

[JaapD]

The ability to remove subject motion or camera shake is much more powerful than what can be done with Focusmagic. I'm still hoping for an upgrade (version 2.x?) that incorporates masking, so the sharpening can be targeted to specific regions of interest.

Cheers,
Bart

Indeed. At least a threshold function, as with USM, would be preferable. Now FM enhances noise in darker areas too much.
Update on rel. 1.3: still too much artefacts, vertical stripes in areas with trees in the image so fully created by Topaz Sharpen (auto function) itself. For me unusable. Wait for a next update. 

Regards,
Jaap.

[Bart_van_der_Wolf]

Indeed. At least a threshold function, as with USM, would be preferable. Now FM enhances noise in darker areas too much.
Update on rel. 1.3: still too much artefacts, vertical stripes in areas with trees in the image so fully created by Topaz Sharpen (auto function) itself. For me unusable. Wait for a next update. 

Yes, it's getting better, but (zoomed in for inspection) not all issues have been fully solved yet.

Cheers,
Bart

[earlybird]

I did some reading about "Machine Learning" today and was fascinated with the concept that the "machine" doesn't necessarily reveal how it has evolved its processes to the code developers.

It seems like optimizing the function requires a bit of input from the coders, but they have to wait to see, and maybe understand, what the machine is doing before they can exert and effective influence.

After using v 1.3 on a batch of pictures the past few days, I am not sure if I think the results have improved when compared to the output of the previous version of Sharpen AI.

I wish there was a practical way to run two or more models on one machine.