From geolocated 2D video to a 4D reconstruction
A post describes converting each frame of geolocated 2D video into 3D representations called Gaussians, placed where and when each moment happened.
TLDR
A post describes reconstructing space and time in 4D by converting geolocated video into per-frame 3D Gaussians and retaining their locations and times. The author wonders whether this resembles seeing “every moment at once,” connecting the idea to Vonnegut’s image of humans as “great millipedes” and suggesting it fits four-legged animals too.
Combined views
—
2 Sources, first seen ago
