Aug3D: Augmenting large scale outdoor datasets for Generalizable Novel View Synthesis

TL;DR

This paper introduces Aug3D, a novel augmentation technique using SfM-reconstructed scenes to improve large-scale outdoor dataset training for generalizable, feed-forward novel view synthesis models, addressing view overlap limitations.

Contribution

The paper proposes Aug3D, a new augmentation method that enhances training data for outdoor NVS models, enabling better generalization and view prediction in large-scale scenes.

Findings

Aug3D improves PSNR by 10% with fewer views per cluster.

Reducing views from 20 to 10 decreases performance, but Aug3D mitigates this.

Aug3D enhances the ability of PixelNeRF to predict novel outdoor views.

Abstract

Recent photorealistic Novel View Synthesis (NVS) advances have increasingly gained attention. However, these approaches remain constrained to small indoor scenes. While optimization-based NVS models have attempted to address this, generalizable feed-forward methods, offering significant advantages, remain underexplored. In this work, we train PixelNeRF, a feed-forward NVS model, on the large-scale UrbanScene3D dataset. We propose four training strategies to cluster and train on this dataset, highlighting that performance is hindered by limited view overlap. To address this, we introduce Aug3D, an augmentation technique that leverages reconstructed scenes using traditional Structure-from-Motion (SfM). Aug3D generates well-conditioned novel views through grid and semantic sampling to enhance feed-forward NVS model learning. Our experiments reveal that reducing the number of views per cluster from 20 to 10 improves PSNR by 10%, but the performance remains suboptimal. Aug3D further addresses this by combining the newly generated novel views with the original dataset, demonstrating its effectiveness in improving the model's ability to predict novel views.