Dan PiponiMay 24, 2025

25 years since MI2.

https://beforesandafters.com/2020/05/27/mission-impossible-ii-a-virtual-production-game-changer/

IIRC I hadn't put my automatically differentiating photogrammetry tool into production at that point. What I did write were some semi-automated tools to align photography with geometry (eg. 3D models of Sydney) and then a completely crazy tool that sliced the models up into pieces according to which cameras they were visible from so each piece could be texture mapped efficiently from that camera.

Show threadPer VognsenMay 24, 2025
@dpiponi Was AD niche back then in computer graphics? It must have been; at least I don't remember it being much of a thing when I was still paying attention to SIGGRAPH in that period (well, a little after).
Show threadDan Piponi

@pervognsen Yeah.

I first found it purely by chance in this paper, but almost nobody in graphics seemed aware of the method: https://dl.acm.org/doi/10.1145/74877.74895

Soon after papers like this appeared: https://grail.cs.washington.edu/projects/control/

But earlier it's hard to find references. For example one use was in Tom Duff's patent for motion blur for blobbies in Renderman.

May 24, 2025 at 12:21amWeb
Show threadCraig KolbMay 24, 2025
@dpiponi @pervognsen The earliest I remember hearing about AD applied to graphics was Don Mitchell's ray tracing system that leveraged AD and interval arithmetic. He had a paper on caustic rendering that used the former. http://mentallandscape.com/Papers_siggraph92.pdf And then of course Tom's AD-based raindrop renderer ;)