Hi,
I was working on some pictures I made yesterday of Hummingbirds in flight. Despite the fact that the birds were in full sunlight, I purposefully choose to underexpose in order to utilize a fast shutter speed, so I knew I would be pushing in post and expected I would use a Denoise process.
I occasionally test Topaz Denoise 6 and Topaz Denoise AI with direct comparisons, and too date I have on each occasion opted to stick with Denoise 6 as the solution of choice.
It had been a while since I launched Denoise AI. Today when I did so I was offered the latest model update, so I was optimistic and eager to try it.
I ran the Topaz Denoise AI v1.2.1 on several pictures using the auto settings, and while the denoising results were ok, I found several patches of vertical linear striping artifacts marring the image. In addition, the automatically chosen sharpening effect was over the top and far more dramatized than I subsequently got with Topaz Sharpen AI in a two step Denoise v6-Sharpen AI process.
I look forward to the next update to Denoise AI, while I contemplate if and how software developers will actually be able to interpret and then effectively influence what the machine learning is teaching to the machine. Will the results of the machine learned processes be based on a grab bag of arbitrary decisions made with less circumspection than traditional algorithmic application?