Report
GAE's geometry-native latent space for 3D-consistent world generation
A post says GAE jointly generates RGB(D), camera poses and geometry, improving visual quality and cross-view consistency.
TLDR
A post describes GAE as using a geometry-native latent space to compress pretrained 3D foundation-model features. It says GAE jointly generates RGB(D), camera poses and geometry, improving visual quality and cross-view consistency.
Combined views
25
1 Source, first seen ago
