Mobile-4DGS: Unified Static-Dynamic Real-time Mobile Gaussian Splatting

Xiaobiao Du     Zheng Fang      Tianqing Zhu      Richard Hartley      Xin Yu
Paper Code

Mobile-4DGS achieves real-time rendering for challenging both dynamic and static scenes on mobile platforms.

Interactive WebGL showcase

Experience Mobile-GS on your phone

Choose a scene, then drag to explore it in real time—right in your browser.

Best experienced in landscape mode on a modern mobile browser.

Abstract

Recent advances in 3D Gaussian Splatting (3DGS) have achieved remarkable performance in novel view synthesis, yet deploying both static and dynamic Gaussian representations on resource-constrained mobile devices remains challenging due to heavy storage, redundant primitives, and costly per-frame computation. We present Mobile-4DGS, a unified lightweight framework for high-fidelity real-time static and dynamic Gaussian rendering on mobile platforms. For compact appearance modeling, we introduce a Monte Carlo Specular Energy Aggregator that compresses high-order radiance residuals into the first-order Spherical Harmonics (SH), together with an Attribute-Conditioned SH Enhancement module whose predicted offsets are pre-baked before inference. We further propose a Multi-View Alpha-Based Densification and Pruning strategy to suppress redundant primitives while maintaining multi-view consistency. For dynamic scenes, we develop a compact explicit 4D representation by constructing second-order Gaussian motion, learnable temporal support, and a binary static–dynamic partition, enabling continuous-time modeling without runtime deformation networks. Based on this partition, a Depth-Order Certificate selectively reuses previously committed depth orders to reduce re-projection, sorting, merging, and index-buffer updates during playback. Extensive experiments on static and dynamic scenes demonstrate that Mobile-4DGS substantially reduces storage and rendering overhead while maintaining competitive visual quality, enabling real-time 3D and 4D Gaussian Splatting on mobile devices.

Motivation

Left: Per Gaussian memory footprint across 3DGS variants. Mobile-4DGS achieves significant compression (61\% and 26\% reductions) by optimizing Spherical Harmonics (SH) coefficients and decoupling SH into the base and view-independent components. Right: Qualitative comparison demonstrates that Mobile-4DGS with only first-order SH can render high-fidelity high-frequency details comparable to 3DGS.

Motivation

Overview of the Mobile-4DGS framework. Our method optimizes third-order SH for the initial 3k iterations, then transitions via Monte Carlo Specular Energy Aggregator for high-frequency representation. With the first-order direction moments inherited from the original third-order SH, we leverage a neural network to aggregate these latents into first-order SH $c'$ for rendering. During inference, the model requires only a one-time decoding step to obtain the first-order SH coefficients, significantly reducing storage while introducing no per-frame decoding overhead.

Implicit Static-dynamic Partion

This video illustrates the central decomposition of our approach on the N3DV flame_steak sequence: the final render is exactly the sum of a dynamic branch and a static branch, dynamic + static = render, yet this partition is never supervised. Each Gaussian carries a learnable gate that is relaxed into a binary decision during training, so the model itself decides whether a Gaussian must move over time or can remain time-invariant. The only driving force is the photometric reconstruction objective, together with a sparsity pressure that encourages as much of the scene as possible to stay static and a binarization pressure that pushes each gate towards a clear decision before the partition is frozen. No segmentation masks, motion labels, or flow supervision are used at any point.