As an Amazon Associate we earn from qualifying purchases.
It's easy to think of StyleSynthesis as a relatively recent development, especially with the explosion of interest in AI art. But looking back, the core ideas have roots in much older experiments. I find it fascinating to trace the lineage, even back to some of the early work in texture synthesis using Markov Random Fields in the late 90s. While the application is wildly diffrent – creating images rather of just analyzing textures – the underlying concept of capturing and re-applying statistical patterns feels remarkably similar.
Thinking about it, even filter effects in early Photoshop versions could be considered rudimentary forms of style transfer. They weren't AI-driven, but they allowed users to impose a pre-defined "style" (like a paint daub effect or a charcoal sketch simulation) onto an image. They lacked the nuance and control we have now, but they were definitely stepping stones.
Does anyone else have examples of early algorithms or approaches that, in retrospect, contributed to the development of modern StyleSynthesis techniques? I'm curious what other seemingly unrelated areas might have played a role.