Is it really bend, or are there LED panels mounted to give something the appearance of being round? This needs to be known. If that is clear, and most of the time, AFAIK, it is based on panels, the work is just based on how savvy one is with UVs.
I think the key is here to have a plane that is bent into place, representing the exact geometry of the target “screen”, and a flat representation that has the exact same polygon set up. In that way, you can project and bake, then use the harvest (copied) UV, and you have it in a flat rectangular way.
In post two, I have shared all single steps with several files.
The key is to know exactly where one should stand to see the effect, as it falls apart when you move away, and to have a good “blueprint” of the video wall, flat, folded, or “bend”. Each is different.
The first step is certainly to get fluent with UVs. Without that, there is no option to understand that. So, since I have no clue if that is new to you or if you’re fluent with this, here is a little introduction drill. Has worked wonders with uncounted people so far.
The technology is moving on, and I believe that the LED Volume, with their real-time engines, will take care of that soon, as they do it already in all these new exciting studios that pop up. I’m not clear how accessible these functions field already, but I don’t see a big secret in it. (It was, after all, part of my final work to get my MFA back in the early ‘90s while creating a VR center, with anything we see these days, virtual studios and anamorphic displays.
One of the best videos I have seen is this here:
So far, I have followed it (with great interest, BTW.), the software in use is set up with the data from the LED Volume (i.e., bent wall), and then the input is calculated based on the tracking of the camera (which represents the sweet spot of these video walls like in Japan for example.
All the best