Next-Scale Autoregressive Models are Zero-Shot Single-Image Object View Synthesizers

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yuan, Shiran, Zhao, Hao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908400107913216
author Yuan, Shiran
Zhao, Hao
author_facet Yuan, Shiran
Zhao, Hao
contents Methods based on diffusion backbones have recently revolutionized novel view synthesis (NVS). However, those models require pretrained 2D diffusion checkpoints (e.g., Stable Diffusion) as the basis for geometrical priors. Since such checkpoints require exorbitant amounts of data and compute to train, this greatly limits the scalability of diffusion-based NVS models. We present Next-Scale Autoregression Conditioned by View (ArchonView), a method that significantly exceeds state-of-the-art methods despite being trained from scratch with 3D rendering data only and no 2D pretraining. We achieve this by incorporating both global (pose-augmented semantics) and local (multi-scale hierarchical encodings) conditioning into a backbone based on the next-scale autoregression paradigm. Our model also exhibits robust performance even for difficult camera poses where previous methods fail, and is several times faster in inference speed compared to diffusion. We experimentally verify that performance scales with model and dataset size, and conduct extensive demonstration of our method's synthesis quality across several tasks. Our code is open-sourced at https://github.com/Shiran-Yuan/ArchonView.
format Preprint
id arxiv_https___arxiv_org_abs_2503_13588
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Next-Scale Autoregressive Models are Zero-Shot Single-Image Object View Synthesizers
Yuan, Shiran
Zhao, Hao
Computer Vision and Pattern Recognition
Graphics
Methods based on diffusion backbones have recently revolutionized novel view synthesis (NVS). However, those models require pretrained 2D diffusion checkpoints (e.g., Stable Diffusion) as the basis for geometrical priors. Since such checkpoints require exorbitant amounts of data and compute to train, this greatly limits the scalability of diffusion-based NVS models. We present Next-Scale Autoregression Conditioned by View (ArchonView), a method that significantly exceeds state-of-the-art methods despite being trained from scratch with 3D rendering data only and no 2D pretraining. We achieve this by incorporating both global (pose-augmented semantics) and local (multi-scale hierarchical encodings) conditioning into a backbone based on the next-scale autoregression paradigm. Our model also exhibits robust performance even for difficult camera poses where previous methods fail, and is several times faster in inference speed compared to diffusion. We experimentally verify that performance scales with model and dataset size, and conduct extensive demonstration of our method's synthesis quality across several tasks. Our code is open-sourced at https://github.com/Shiran-Yuan/ArchonView.
title Next-Scale Autoregressive Models are Zero-Shot Single-Image Object View Synthesizers
topic Computer Vision and Pattern Recognition
Graphics
url https://arxiv.org/abs/2503.13588