MAC-Splat: Multi-Attribute Consistency for High-Fidelity Sparse-View Reconstruction

2026-07-12

AI Summary

This work addresses the challenge of high-fidelity 3D reconstruction from sparse views, which is often plagued by geometric artifacts due to ambiguities in depth and correspondence when relying solely on 2D photometric losses. To mitigate this, the authors propose MAC-Splat, a framework built upon 3D Gaussian Splatting that integrates MASt3R-based geometric estimates with frozen DINOv3 semantic features to formulate a multi-attribute consistency (MAC) loss in world coordinates. This loss jointly regularizes the position, shape, and appearance of corresponding Gaussians, introducing for the first time direct 3D multi-attribute consistency supervision to significantly reduce reconstruction ambiguity under sparse input views. Evaluated on ScanNet++, MAC-Splat achieves a PSNR gain of over 4.5 dB and lower LPIPS compared to Splatt3R, while maintaining robust performance even with large viewpoint intervals.

Abstract

Reconstructing high-fidelity 3D scenes from sparse-views remains a central problem in generalizable neural rendering. Existing generalizable 3D Gaussian Splatting (3DGS) methods often exhibit geometric artifacts in sparse-view settings, since supervision based solely on 2D photometric losses cannot resolve depth and correspondence ambiguities. To address this issue, we propose MAC-Splat, a training framework built around direct 3D consistency supervision. MAC-Splat builds on the MASt3R geometric backbone and a frozen DINOv3 encoder to obtain semantically informed 2D correspondences, which serve as geometric anchors for 3D supervision. Using these anchors, we define the Multi-Attribute Consistency (MAC) loss. This objective jointly regularizes the 3D attributes of matched Gaussians, including their position, shape, and appearance, by enforcing agreement in a common world coordinate frame. The formulation is robust to outliers and respects the geometry of covariance matrices, which leads to stable training under sparse-view conditions. Experiments on ScanNet++ show that MAC-Splat outperforms strong baselines, with particularly large gains under different overlap regimes. In particular, it improves average PSNR over Splatt3R by more than 4.5 dB, reduces LPIPS, and maintains performance as the camera pose gap increases. These results indicate that a direct, multi-attribute 3D consistency objective, when combined with high-quality correspondences, is effective for addressing the ill-posed sparse-view reconstruction problem.

Problem

• sparse-view reconstruction
• 3D Gaussian Splatting
• geometric ambiguity
• high-fidelity 3D reconstruction
• correspondence ambiguity

Innovation

• Multi-Attribute Consistency
• 3D Gaussian Splatting
• Sparse-View Reconstruction
• 3D Consistency Supervision
• DINOv3