# ${C}^{3}$-GS: Learning Context-aware, Cross-dimension, Cross-scale Feature for Generalizable Gaussian Splatting

## TL;DR

C³-GS introduces a novel framework that enhances feature learning for Gaussian Splatting by incorporating context-aware, cross-dimension, and cross-scale constraints, leading to improved generalization and photorealistic novel view synthesis.

## Contribution

It proposes a new architecture with three lightweight modules that improve feature fusion and generalization in Gaussian Splatting without extra supervision.

## Key findings

- Achieves state-of-the-art rendering quality on benchmark datasets.
- Demonstrates superior generalization to unseen scenes.
- Produces photorealistic images from sparse input views.

## Abstract

Generalizable Gaussian Splatting aims to synthesize novel views for unseen scenes without per-scene optimization. In particular, recent advancements utilize feed-forward networks to predict per-pixel Gaussian parameters, enabling high-quality synthesis from sparse input views. However, existing approaches fall short in encoding discriminative, multi-view consistent features for Gaussian predictions, which struggle to construct accurate geometry with sparse views. To address this, we propose $\mathbf{C}^{3}$-GS, a framework that enhances feature learning by incorporating context-aware, cross-dimension, and cross-scale constraints. Our architecture integrates three lightweight modules into a unified rendering pipeline, improving feature fusion and enabling photorealistic synthesis without requiring additional supervision. Extensive experiments on benchmark datasets validate that $\mathbf{C}^{3}$-GS achieves state-of-the-art rendering quality and generalization ability. Code is available at: https://github.com/YuhsiHu/C3-GS.