there’s been a lot of work done on that in the last decade or so, the current algorithms give pretty good results? was amazed to fiddle with Blender’s camera tracking a few years ago
edit turns out this tutorial is by the VFX guy for Prospect & the Dynamo Dream clips and he has a verbal presentation that’s way more like Pauly Shore than I would have expected
what does this mean? like a 3D model? or just another photorealistic image? the video seemed to show only additional images, nothing 3D rendered. sorry I’m not high anymore but even more confused now.
for their road video example, I think they’ve used a reference video (that was used for vid2vid) and extracted sematic tags with computer vision techniques
it’s kind of wild, as game players who think about 3D spaces a lot, to find graphics research that gives absolutely zero shits about model extraction and only care about reasonable-looking outputs or ease of use
I use public-facing achievements/trophy data to look at competitor products all the time – keep up a sense of how long players stuck around by comparing the standard 1/3rd completion rate, look for unexpectedly popular or unpopular activities signified by achievements and the like. For our own stuff I’ve always had or implemented my own telemetry; at a studio of 150+ people you’ll always find a data analyst doing roughly the same work as you’d find at any web business.
did you know that you can enable two factor authentication practically everywhere, including for this very web forum?
This advice is useless for libretro! my sympathies go out to them as they work out how to restore their repos from locally-cached copies/automatic GitHub archives for deleted repos
edit oh no, they did have 2FA enabled but only for almost everyone
Can’t seem to find out much info. as to why somebody would do this? There are references on twitter to this having happened before. Anyone know the full story here?