Spackle: Completing Large View Single Image NVS with Adaptive Gaussians
cs.CV, cs.AI
Submitted: 2026-09-25
Updated: 2026-09-25
License: http://creativecommons.org/licenses/by/4.0/
The gist: Single-image novel view synthesis (NVS) enables photorealistic rendering of un- observed viewpoints from a single input.
Terminology
Abstract
Single-image novel view synthesis (NVS) enables photorealistic rendering of un- observed viewpoints from a single input. Practical NVS systems require two key capabilities: robust reconstruction of occluded regions and high inference effi- ciency. While hybrid decoupled frameworks combining feedforward 3D Gaussian Splatting (3DGS) and diffusion models show promise for large-view-deviation NVS, they suffer from capacity competition: a fixed number of Gaussians forces resource shifts from visible to newly disoccluded areas, degrading original scene fidelity when the target view deviates significantly from the input. To address this, we propose Spackle, a lightweight residual learning framework that mit- igates capacity competition without sacrificing efficiency. Spackle operates in three stages: predicting base 3DGS attributes from given views, automatically identifying poorly reconstructed regions, and learning a residual 3DGS optimized exclusively for these areas. At inference, we combine the baseline and aug- mented Gaussians for NVS. We conduct comprehensive experiments and show that Spackle achieves state-of-the-art performance on large-view-deviation cases.
Sources
- Lyra: Generative 3D Scene Reconstruction via Video Diffusion Model Self-Distillation
- Image Quality Assessment: Unifying Structure and Texture Similarity
- NavCrafter: Exploring 3D Scenes from a Single Image
- CAT3D: Create Anything in 3D with Multi-View Diffusion Models
- CausNVS: Autoregressive Multi-view Diffusion for Flexible 3D Novel View Synthesis
- SyncFix: Fixing 3D Reconstructions via Multi-View Synchronization
- FlashWorld: High-quality 3D Scene Generation within Seconds
- Sharp Monocular View Synthesis in Less Than a Second
- One View Is Enough: In-the-Wild Monocular Pretraining for Novel View Generation
- Lyra 2.0: Explorable Generative 3D Worlds
- DreamGaussian: Generative Gaussian Splatting for Efficient 3D Content Creation
- One2Scene: Geometric Consistent Explorable 3D Scene Generation from a Single Image
- ViewCrafter: Taming Video Diffusion Models for High-fidelity Novel View Synthesis
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