Autoregressive latent generator turns multi-view images into 3D Gaussian scenes
kwangmoo_yi · x · 2026-09-05
Lucas and Pietrantoni et al. present "Sparse auto-regressive modeling for scene generation from multi-view images": multi-view images plus pointmaps feed an autoregressive latent-space generator, decoded as 3D Gaussians for scene generation.
Related event: SPAR3S Generates Full 3D Scenes from Sparse Views Without 3D Supervision(2 posts)→
More from Multimodal
- VeriPhy: agentic physical reasoning framework for world model evaluation — Wenzhuo Xu · 2026-09-05
- Grok Imagine Odyssey Contest Winners Announced by xAI — Mr_AllenT · 2026-09-05
- 3D Reconstruction Studio LichtFeld Studio Shuts Down — janusch_patas · 2026-09-05
- Two hours from scratch: building playable creations in Flora AI — round · 2026-09-05
- AI-made static headers become fully interactive: 'we're so early on the future of interfaces' — round · 2026-09-05
- Asking Astra to open MS Paint and draw you: 'We're living through a golden age' — idanbeck · 2026-09-05