GIFSplat: Generative Prior-Guided Iterative Feed-Forward 3D Gaussian Splatting from Sparse Views

TL;DR

GIFSplat is a novel feed-forward iterative framework for 3D Gaussian Splatting from sparse views, integrating generative priors to improve reconstruction quality efficiently without test-time optimization.

Contribution

It introduces a purely feed-forward iterative refinement method that incorporates generative priors via distillation, enhancing 3D reconstruction from sparse views while maintaining efficiency.

Findings

Outperforms state-of-the-art feed-forward methods in PSNR by up to +2.1 dB.

Maintains second-scale inference time without camera poses or gradient optimization.

Effective across multiple datasets including DL3DV, RealEstate10K, and DTU.

Abstract

Feed-forward 3D reconstruction offers substantial runtime advantages over per-scene optimization, which remains slow at inference and often fragile under sparse views. However, existing feed-forward methods still have potential for further performance gains, especially for out-of-domain data, and struggle to retain second-level inference time once a generative prior is introduced. These limitations stem from the one-shot prediction paradigm in existing feed-forward pipeline: models are strictly bounded by capacity, lack inference-time refinement, and are ill-suited for continuously injecting generative priors. We introduce GIFSplat, a purely feed-forward iterative refinement framework for 3D Gaussian Splatting from sparse unposed views. A small number of forward-only residual updates progressively refine current 3D scene using rendering evidence, achieve favorable balance between efficiency and quality. Furthermore, we distill a frozen diffusion prior into Gaussian-level cues from enhanced novel renderings without gradient backpropagation or ever-increasing view-set expansion, thereby enabling per-scene adaptation with generative prior while preserving feed-forward efficiency. Across DL3DV, RealEstate10K, and DTU, GIFSplat consistently outperforms state-of-the-art feed-forward baselines, improving PSNR by up to +2.1 dB, and it maintains second-scale inference time without requiring camera poses or any test-time gradient optimization.