PixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction

By David Charatan, Sizhe Li, Andrea Tagliasacchi, Vincent Sitzmann. Computer Vision and Pattern Recognition, 2023

744 citations recorded by the source corpus.

We introduce pixelSplat, a feed-forward model that learns to reconstruct 3D radiance fields parameterized by 3D Gaussian primitives from pairs of images. Our model features real-time and memory-efficient rendering for scalable training as well as fast 3D reconstruction at inference time. To overcome local minima inherent to sparse and locally supported representations, we predict a dense probability distribution over 3D and sample Gaussian means from that probability distribution. We make this sampling operation differentiable via a reparameterization trick, allowing us to back-propagate gradients through the Gaussian splatting representation. We benchmark our method on wide-baseline novel view synthesis on the real-world RealEstate10k and ACID datasets, where we outperform state-of-the-art light field transformers and accelerate rendering by 2.5 orders of magnitude while reconstructing an interpretable and ed-itable 3D radiance field. Additional materials can be found on the project website.11dcharatan.github.io/pixelsplat

Subfields: Computational Mechanics, Computer Graphics and Computer-Aided Design, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/632a4956aebedd85f61445718ec5df9589bb5843

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Mechanics

In Computer Graphics and Computer-Aided Design

In Computer Vision and Pattern Recognition

Citations in the corpus

Gen3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control (2025, positive): “We compare our method to two representative works for sparse view reconstruction, Pixel-Splat [6] and MVSplat [8].”
GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting (2024, methodology): “For scene-level, we adopt two input views for a fair comparison with pixelSplat [8].”
LGM: Large Multi-View Gaussian Model for High-Resolution 3D Content Creation (2024, background): “Similarly, PixelSplat [3] predicts Gaussian parameters for each pixel of two posed images from scene datasets.”
ViewCrafter: Taming Video Diffusion Models for High-fidelity Novel View Synthesis (2024, methodology): “Inspired by the success of 3D-GS [2], recent approaches such as PixelSplat [27] and MVSplat [28] explore training regression-based models to produce 3D Gaussian representations for real-time rendering capabilities.”
MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Images (2024, methodology): “Recently, there have been growing interests [1,3,5,12,26,36,39,43,44,51] in scene reconstruction and synthesis from sparse input views ( e.g ., 2 or 3).”
Depth Anything 3: Recovering the Visual Space from Any Views (2025, positive): “Follow-up models extended this paradigm with multi-view inputs [10, 85, 94, 110], video input [19, 59, 94, 121], robust correspondence modeling [48], camera parameter injection [39, 43], large-scale SfM [18], SLAM applications [54], and view synthesis with 3D Gaussians [11, 13, 41, 79, 108, 122].”