BRUM: Robust 3D Vehicle Reconstruction from 360 Sparse Images

TL;DR

This paper introduces BRUM, a robust method for 3D vehicle reconstruction from sparse images, combining enhanced Gaussian Splatting, selective photometric loss, and improved pose estimation to achieve state-of-the-art results.

Contribution

The paper presents a novel approach that improves 3D vehicle reconstruction from sparse views by integrating a new pose estimation architecture and a selective loss function.

Findings

Achieves high-quality reconstructions with sparse input views.

Outperforms existing methods on multiple benchmarks.

Introduces a new dataset for vehicle reconstruction evaluation.

Abstract

Accurate 3D reconstruction of vehicles is vital for applications such as vehicle inspection, predictive maintenance, and urban planning. Existing methods like Neural Radiance Fields and Gaussian Splatting have shown impressive results but remain limited by their reliance on dense input views, which hinders real-world applicability. This paper addresses the challenge of reconstructing vehicles from sparse-view inputs, leveraging depth maps and a robust pose estimation architecture to synthesize novel views and augment training data. Specifically, we enhance Gaussian Splatting by integrating a selective photometric loss, applied only to high-confidence pixels, and replacing standard Structure-from-Motion pipelines with the DUSt3R architecture to improve camera pose estimation. Furthermore, we present a novel dataset featuring both synthetic and real-world public transportation vehicles, enabling extensive evaluation of our approach. Experimental results demonstrate state-of-the-art performance across multiple benchmarks, showcasing the method's ability to achieve high-quality reconstructions even under constrained input conditions.