SR3R: Rethinking Super-Resolution 3D Reconstruction With Feed-Forward Gaussian Splatting

TL;DR

SR3R introduces a feed-forward approach for 3D super-resolution that learns directly from large-scale multi-scene data, enabling high-fidelity, real-time 3D scene reconstruction with strong generalization to unseen scenes.

Contribution

The paper proposes SR3R, a novel feed-forward framework that predicts high-resolution 3D Gaussian Splatting representations directly from sparse low-resolution views, improving fidelity and generalization.

Findings

Outperforms state-of-the-art 3DSR methods on multiple benchmarks.

Achieves strong zero-shot generalization to unseen scenes.

Enables real-time 3D reconstruction with high fidelity.

Abstract

3D super-resolution (3DSR) aims to reconstruct high-resolution (HR) 3D scenes from low-resolution (LR) multi-view images. Existing methods rely on dense LR inputs and per-scene optimization, which restricts the high-frequency priors for constructing HR 3D Gaussian Splatting (3DGS) to those inherited from pretrained 2D super-resolution (2DSR) models. This severely limits reconstruction fidelity, cross-scene generalization, and real-time usability. We propose to reformulate 3DSR as a direct feed-forward mapping from sparse LR views to HR 3DGS representations, enabling the model to autonomously learn 3D-specific high-frequency geometry and appearance from large-scale, multi-scene data. This fundamentally changes how 3DSR acquires high-frequency knowledge and enables robust generalization to unseen scenes. Specifically, we introduce SR3R, a feed-forward framework that directly predicts HR 3DGS representations from sparse LR views via the learned mapping network. To further enhance reconstruction fidelity, we introduce Gaussian offset learning and feature refinement, which stabilize reconstruction and sharpen high-frequency details. SR3R is plug-and-play and can be paired with any feed-forward 3DGS reconstruction backbone: the backbone provides an LR 3DGS scaffold, and SR3R upscales it to an HR 3DGS. Extensive experiments across three 3D benchmarks demonstrate that SR3R surpasses state-of-the-art (SOTA) 3DSR methods and achieves strong zero-shot generalization, even outperforming SOTA per-scene optimization methods on unseen scenes.