Fast 3D Foundation Model Initialized Gaussian Splatting

Summary · qwen2.5:32b

This paper presents a fast method for high-quality 3D Gaussian Splatting reconstruction that bypasses traditional Structure-from-Motion techniques by using 3D Foundation Models for initialization, achieving competitive results (23.61 dB PSNR, 0.19 LPIPS) in about three minutes per scene with as few as 50-60 input views.

Suggested post angle

Fast 3D rendering for robotics, VR, and autonomous navigation applications.

Excerpt

arXiv:2607.03209v1 Announce Type: cross Abstract: This paper introduces a fast method for high-quality 3D Gaussian Splatting (3DGS) reconstruction without traditional Structure-from-Motion (SfM). The proposed approach leverages 3D Foundation Models (3DFMs) for camera pose and point-cloud initialization, then jointly optimizes both camera poses and Gaussian primitives using a depth-guided loss function. This enables fast convergence even from rough initialization with as few as 50-60 input views. To further improve reconstruction quality in sparse-view scenarios, an MLP-based pose refinement module is introduced alongside depth-guided supervision from the foundation model. Extensive experiments on Mip-NeRF 360, Tanks and Temples, and RobustNeRF demonstrate that the proposed method achieves competitive reconstruction quality (23.61 dB PSNR, 0.19 LPIPS) while reducing training time to approximately three minutes per scene. The proposed method produces ready-to-use 3DGS models at a fraction of the time required by existing pipelines, making it suitable for near real-time applications in robotics, VR, and autonomous navigation.