IDESplat: Iterative Depth Probability Estimation for Generalizable 3D Gaussian Splatting

TL;DR

IDESplat tackles the bottleneck of Gaussian mean prediction in generalizable 3D Gaussian Splatting by iteratively boosting depth probability through cascaded Warp-Index Epipolar Attention layers (DPBUs) and a multiplicative fusion strategy. A Gaussian Focused Module further refines remaining Gaussian parameters via sparsified, layer-wise attention over salient tokens. The approach achieves state-of-the-art reconstruction quality and strong cross-dataset generalization on RealEstate10K, ACID, and DTU with significantly fewer parameters and memory than prior methods. By combining iterative depth probability estimation with memory-efficient cross-view matching, IDESplat enables real-time 3D scene reconstruction while maintaining high fidelity and generalization.

Abstract

Generalizable 3D Gaussian Splatting aims to directly predict Gaussian parameters using a feed-forward network for scene reconstruction. Among these parameters, Gaussian means are particularly difficult to predict, so depth is usually estimated first and then unprojected to obtain the Gaussian sphere centers. Existing methods typically rely solely on a single warp to estimate depth probability, which hinders their ability to fully leverage cross-view geometric cues, resulting in unstable and coarse depth maps. To address this limitation, we propose IDESplat, which iteratively applies warp operations to boost depth probability estimation for accurate Gaussian mean prediction. First, to eliminate the inherent instability of a single warp, we introduce a Depth Probability Boosting Unit (DPBU) that integrates epipolar attention maps produced by cascading warp operations in a multiplicative manner. Next, we construct an iterative depth estimation process by stacking multiple DPBUs, progressively identifying potential depth candidates with high likelihood. As IDESplat iteratively boosts depth probability estimates and updates the depth candidates, the depth map is gradually refined, resulting in accurate Gaussian means. We conduct experiments on RealEstate10K, ACID, and DL3DV. IDESplat achieves outstanding reconstruction quality and state-of-the-art performance with real-time efficiency. On RE10K, it outperforms DepthSplat by 0.33 dB in PSNR, using only 10.7% of the parameters and 70% of the memory. Additionally, our IDESplat improves PSNR by 2.95 dB over DepthSplat on the DTU dataset in cross-dataset experiments, demonstrating its strong generalization ability.

Figures (11)

• Figure 1: Left: Methods chen2024mvsplatxu2025depthsplatliu2025monosplat that estimate depth probability via a single warp operation. Middle: Our IDESplat can iteratively leverage multi-warp operations to boost the depth probability estimate and refine the depth candidates for accurate Gaussian mean predictions. Right: The experimental results of our IDESplat compared with mainstream methods such as PixelSplat charatan2024pixelsplat, MVSplat chen2024mvsplat, MonoSplat liu2025monosplat, and DepthSplat xu2025depthsplat. The PSNR values are reported for the entire RE10K test set.
• Figure 2: The overall architecture of IDESplat. IDESplat is composed of three key parts: a feature extraction backbone, an iterative depth probability estimation process, and a Gaussian Focused Module (GFM). The iterative process consists of cascaded Depth Probability Boosting Units (DPBUs). Each unit combines multi-level warp results in a multiplicative manner to mitigate the inherent instability of a single warp. As IDESplat iteratively updates the depth candidates and boosts the probability estimates, the depth map becomes more precise, leading to accurate Gaussian means.
• Figure 3: The comparison of visualization results for novel view synthesis on the RealEstate10K dataset. Our IDESplat significantly outperforms previous state-of-the-art methods in rendering challenging regions.
• Figure 4: Comparison of depth prediction maps for different models on the RE10K dataset.
• Figure 5: The architecture of IDESplat. IDESplat comprises a feature extraction backbone, an iterative depth estimation process with cascaded Depth Probability Boosting Units (DPBUs), and a Gaussian Focused Module (GFM).