LVSPM: Long Sequence View Synthesis and Pose Estimation Model
cs.CV
Submitted: 2026-10-07
Updated: 2026-10-07
Comments: ECCV 2026. Project Page: https://burningdust21.github.io/Projects/LVSPM/
Project page: https://burningdust21.github.io/Projects/LVSPM
License: http://creativecommons.org/licenses/by-sa/4.0/
The gist: We present LVSPM, a generalizable model that jointly estimates camera poses and synthesizes novel views from uncalibrated image collections.
Terminology
Abstract
We present LVSPM, a generalizable model that jointly estimates camera poses and synthesizes novel views from uncalibrated image collections. Trained with only RGB images and pose supervision, LVSPM avoids dense 3D ground truth and employs test-time training (TTT) layers to scale seamlessly to hundreds of input views. On RealEstate10k, Co3Dv2, and DL3DV, LVSPM surpasses VGGT in pose estimation across 16-256 views, with especially large margins at strict thresholds. For novel view synthesis under a practical protocol where more views cover larger scenes, LVSPM achieves state-of-the-art pose-free quality---surpassing even pose-dependent models in PSNR---and still maintains high quality as scene scale grows, while baselines collapse. The code is available at https://burningdust21.github.io/Projects/LVSPM.
Sources
- TTT3R: 3D Reconstruction as Test-Time Training
- Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality
- InstantSplat: Sparse-view Gaussian Splatting in Seconds
- Mamba: Linear-Time Sequence Modeling with Selective State Spaces
- No Pose at All: Self-Supervised Pose-Free 3D Gaussian Splatting from Sparse Views
- LEAP: Liberate Sparse-view 3D Modeling from Camera Poses
- RayZer: A Self-supervised Large View Synthesis Model
- AnySplat: Feed-forward 3D Gaussian Splatting from Unconstrained Views
- Few-View Object Reconstruction with Unknown Categories and Camera Poses
- RelPose++: Recovering 6D Poses from Sparse-view Observations
- Depth Anything 3: Recovering the Visual Space from Any Views
- MVSGaussian: Fast Generalizable Gaussian Splatting Reconstruction from Multi-View Stereo
- R2D2: Repeatable and Reliable Detector and Descriptor
- GLU Variants Improve Transformer
- Splatt3R: Zero-shot Gaussian Splatting from Uncalibrated Image Pairs
- Learning to (Learn at Test Time): RNNs with Expressive Hidden States
- MeshLRM: Large Reconstruction Model for High-Quality Meshes
- SiNeRF: Sinusoidal Neural Radiance Fields for Joint Pose Estimation and Scene Reconstruction
- FreeSplatter: Pose-free Gaussian Splatting for Sparse-view 3D Reconstruction
- No Pose, No Problem: Surprisingly Simple 3D Gaussian Splats from Sparse Unposed Images
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models