arXiv:2511.14149v1 Announce Type: new 
Abstract: Recent trends in SLAM and visual navigation have embraced 3D Gaussians as the preferred scene representation, highlighting the importance of estimating camera poses from a single image using a pre-built Gaussian model. However, existing approaches typically rely on an iterative \textit{render-compare-refine} loop, where candidate views are first rendered using NeRF or Gaussian Splatting, then compared against the target image, and finally, discrepancies are used to update the pose. This multi-round process incurs significant computational overhead, hindering real-time performance in robotics. In this paper, we propose iGaussian, a two-stage feed-forward framework that achieves real-time camera pose estimation through direct 3D Gaussian inversion. Our method first regresses a coarse 6DoF pose using a Gaussian Scene Prior-based Pose Regression Network with spatial uniform sampling and guided attention mechanisms, then refines it through feature matching and multi-model fusion. The key contribution lies in our cross-correlation module that aligns image embeddings with 3D Gaussian attributes without differentiable rendering, coupled with a Weighted Multiview Predictor that fuses features from Multiple strategically sampled viewpoints. Experimental results on the NeRF Synthetic, Mip-NeRF 360, and T\&T+DB datasets demonstrate a significant performance improvement over previous methods, reducing median rotation errors to 0.2{\deg} while achieving 2.87 FPS tracking on mobile robots, which is an impressive 10 times speedup compared to optimization-based approaches. Code: https://github.com/pythongod-exe/iGaussian

تقدم الورقة iGaussian، وهو إطار جديد لتقدير وضع الكاميرا في الوقت الحقيقي باستخدام نهج التغذية الأمامية القائم على عكس التوزيع الغاوسي ثلاثي الأبعاد. تعالج هذه الطريقة القيود المفروضة على العمليات التكرارية التقليدية في الملاحة البصرية وSLAM، والتي غالبًا ما تتطلب عبئًا حسابيًا كبيرًا. من خلال استخدام إطار من مرحلتين، تقوم iGaussian أولاً بتقدير وضع 6DoF تقريبي ثم تقوم بتحسينه، مما يسمح بتقدير أسرع وأكثر كفاءة لوضع الكاميرا من صورة واحدة.

El artículo presenta iGaussian, un nuevo marco para la estimación de la pose de la cámara en tiempo real mediante un enfoque feed-forward basado en la inversión de splatting gaussiano 3D. Este método aborda las limitaciones de los procesos iterativos tradicionales en la navegación visual y SLAM, que a menudo implican una sobrecarga computacional significativa. Al emplear un marco de dos etapas, iGaussian primero estima una pose 6DoF gruesa y luego la refina, permitiendo una estimación de pose de cámara más rápida y eficiente a partir de una sola imagen.

L'article présente iGaussian, un nouveau cadre pour l'estimation de pose de caméra en temps réel utilisant une approche feed-forward basée sur l'inversion de splatting gaussien 3D. Cette méthode répond aux limitations des processus itératifs traditionnels dans la navigation visuelle et le SLAM, qui impliquent souvent une surcharge computationnelle significative. En employant un cadre en deux étapes, iGaussian estime d'abord une pose 6DoF grossière, puis la raffine, permettant une estimation de pose de caméra plus rapide et efficace à partir d'une seule image.

The paper introduces iGaussian, a novel framework for real-time camera pose estimation using a feed-forward approach based on 3D Gaussian splatting inversion. This method addresses the limitations of traditional iterative processes in visual navigation and SLAM, which often involve significant computational overhead. By employing a two-stage framework, iGaussian first estimates a coarse 6DoF pose and then refines it, enabling faster and more efficient camera pose estimation from a single image.

iGaussian: Real-Time Camera Pose Estimation via Feed-Forward 3D Gaussian Splatting Inversion

Was this article worth reading? Share it

Uwear

Deptho.ai

Video Face Swap AI — Discover what people think of this product.