Background for context:
Recently bought a lume pad gen 1. experimenting with the new depth map tech called depth anything. On GitHub it is supposed to be better than midas and supports tensor cores on nvidia gpus, results are decently fast for what it’s doing generating a depth map for 2d video. Like a few minutes for a trailer or a few hours for a movie on a rtx 3060. From there I’m using a python script for stereo matching to make 3d sbs video conversions. I can share the script if there is interest in it. Made it with copilot to help me with it lol.
Anyway my question is how are the 2x2 videos formatted? 2d video in top left offset video horizontally in top right? Offset downward from top left in bottom left? I’ve been enjoying watching 3d sbs trailers using it but though I could mod the script for 2x2?
The Quad format has been sunset and is no longer supported.
However, the format is simple, similar to an SBS, but instead of two horizontal views next to each other, it’s 4 horizontal views with two views stacked on top of the other two.
Views are captured / positioned in the following way:
1 2 3 4
But in the Quad format, the 2x2 is laid out like what’s shown below as a single frame:
1 2
3 4
Just make sure _2x2 is at the end of each filename.