Topaz is a small company, and I once wrote that we need to support such companies to be competitive with large corporations like Adobe.
Here’s my next idea: Many advanced programmers are also graphic designers. This is not a rule, but it happens. Topaz could buy ready-made, taught neural models based on PyTorch or TensowFlow (commonly known as AI filters) from these developers. Or Topaz could pay them a percentage of the sale.
Currently, Topaz has a lot of interesting AI filters such as Gigapixel, Sharpen, Denoise, Jpg2Raw and AI Remix. However, this is still a very small offer, because Topaz cannot do everything alone.
My proposal will cause that many new, unique AI filters will appear quickly, which can then be commercialized.
BTW: For example, I am working on art filters that can make a sketch from photographs, without artifacts, and at a much better level than AI Remix or Topaz Impression. Similarly, you could do painting filters or removing selected elements from the image, doing retouching and other things.
I would like you users and you developers from Topaz to consider this proposal, because it can cause a significant development of the graphics market associated with the Topaz Labs products and the need for more filters.
Especially during the coronavirus, when more and more people will work at home.
Topaz should only develop some technical standard of neural models for external developers. I understand that Topaz will not describe its algorithms due to the market and other companies. This is a secret and know-how. However, the trained models should be about the same standard, or Topaz should develop model translators from one system to its system.
For example, I prefer to use PyTorch standard models because they are easily programmed and more convenient to develop than TensorFlow. Although I also know TensorFlow, my system likes PyTorch more so I prefer to use it.
So the point is for Topaz to be able to import different models from outsiders and then place them in its own products (Topaz uses TensorFlow). Or Topaz could also develop its own frameworks to include in its own products, but working with PyTorch.
Of course, this is only a sketch, a draft, a proposal. You need to add details.
! There may be another way: Topaz could write a program for everyone who wants them to train their own AI models at home and then sell them to Topaz for a percentage.
In this way, Topaz’s offer could be enriched with things that are not there now. And the number of applications would start to increase and compete with Adobe applications and other companies, which would be good from the point of view of end users, i.e. graphic designers and photographers, but also amateurs.
Regards,
Lech Balcerzak
PS
But I do not only mean myself, but also others. I’m sure there are many outside developers who have ideas for new AI filters and would like to do something and then sell it with Topaz Labs.
