The authors frame view synthesis and pose estimation as a single sequence-to-sequence token prediction task.
Joint pose estimation and view synthesis scale to 256 unposed images without 3D labels
LVSPM trains only on RGB images and camera poses, yet beats geometry-supervised rivals in pose accuracy and pose-dependent baselines in rendering quality.
Industry
Xi Chen · Yachi Zhang · Linghao Chen · Minghua Liu · Hao Su · Zexiang Xu · +1 more
UC San Diego · Sudo AI GmbH
Research Digest··3 min read
The authors present LVSPM, a feed-forward model that jointly estimates camera poses and renders novel views from long, unposed image sequences.
Why this paper
From UC San Diego and Sudo AI GmbH
In one line
LVSPM jointly estimates camera poses and synthesizes novel views from unposed image sequences, scaling to 256 views using only RGB and pose supervision.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§