Gaussian Splatting is an Effective Data Generator for 3D Object Detection

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zanjani, Farhad G., Abati, Davide, Wiggers, Auke, Kalatzis, Dimitris, Petersen, Jens, Cai, Hong, Habibian, Amirhossein
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913805905166336
author Zanjani, Farhad G.
Abati, Davide
Wiggers, Auke
Kalatzis, Dimitris
Petersen, Jens
Cai, Hong
Habibian, Amirhossein
author_facet Zanjani, Farhad G.
Abati, Davide
Wiggers, Auke
Kalatzis, Dimitris
Petersen, Jens
Cai, Hong
Habibian, Amirhossein
contents We investigate data augmentation for 3D object detection in autonomous driving. We utilize recent advancements in 3D reconstruction based on Gaussian Splatting for 3D object placement in driving scenes. Unlike existing diffusion-based methods that synthesize images conditioned on BEV layouts, our approach places 3D objects directly in the reconstructed 3D space with explicitly imposed geometric transformations. This ensures both the physical plausibility of object placement and highly accurate 3D pose and position annotations. Our experiments demonstrate that even by integrating a limited number of external 3D objects into real scenes, the augmented data significantly enhances 3D object detection performance and outperforms existing diffusion-based 3D augmentation for object detection. Extensive testing on the nuScenes dataset reveals that imposing high geometric diversity in object placement has a greater impact compared to the appearance diversity of objects. Additionally, we show that generating hard examples, either by maximizing detection loss or imposing high visual occlusion in camera images, does not lead to more efficient 3D data augmentation for camera-based 3D object detection in autonomous driving.
format Preprint
id arxiv_https___arxiv_org_abs_2504_16740
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Gaussian Splatting is an Effective Data Generator for 3D Object Detection
Zanjani, Farhad G.
Abati, Davide
Wiggers, Auke
Kalatzis, Dimitris
Petersen, Jens
Cai, Hong
Habibian, Amirhossein
Computer Vision and Pattern Recognition
We investigate data augmentation for 3D object detection in autonomous driving. We utilize recent advancements in 3D reconstruction based on Gaussian Splatting for 3D object placement in driving scenes. Unlike existing diffusion-based methods that synthesize images conditioned on BEV layouts, our approach places 3D objects directly in the reconstructed 3D space with explicitly imposed geometric transformations. This ensures both the physical plausibility of object placement and highly accurate 3D pose and position annotations. Our experiments demonstrate that even by integrating a limited number of external 3D objects into real scenes, the augmented data significantly enhances 3D object detection performance and outperforms existing diffusion-based 3D augmentation for object detection. Extensive testing on the nuScenes dataset reveals that imposing high geometric diversity in object placement has a greater impact compared to the appearance diversity of objects. Additionally, we show that generating hard examples, either by maximizing detection loss or imposing high visual occlusion in camera images, does not lead to more efficient 3D data augmentation for camera-based 3D object detection in autonomous driving.
title Gaussian Splatting is an Effective Data Generator for 3D Object Detection
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2504.16740