READ: Large-Scale Neural Scene Rendering for Autonomous Driving

Li, Zhuopeng; Li, Lü; Ma, Zeyu; Zhang, Ping; Chen, Junbo; Zhu, Jianke

doi:10.48550/arxiv.2205.05509

Cited by 2 publications

(4 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…NeRF [27] is a powerful tool for novel view synthesis, which represents a scene with a fully connected neural network and optimizes it with differentiable volume rendering. It has been recently applied to autonomous driving scenarios [43,24,29,13]. Block-NeRF [43] reconstructs a whole city by merging multiple block-NeRFs with predicted visibility.…”

Section: Related Workmentioning

confidence: 99%

“…To overcome this, voxel-based NeRF [12,41,30] and depthsupervised NeRF [9], which are adopted in our method, have been proposed to accelerate the training of NeRF. Moreover, some works begin to explore how to edit scenes with NeRF [31,21,24] for autonomous driving. However, these works are not extended to augmenting the data for 3D perception tasks.…”

Section: Related Workmentioning

confidence: 99%

“…PNF [21] does not consider driving data only has few views for most objects, which restricts the quality of novel view synthesis. READ [24] realizes autonomous driving scene editing with larger viewpoints of objects by point-based NeRF. However, this method relies on LiDAR data which limits its application scope.…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

3D Data Augmentation for Driving Scenes on Camera

Tong¹,

Xie²,

Li³

et al. 2023

Preprint

View full text Add to dashboard Cite

Driving scenes are extremely diverse and complicated that it is impossible to collect all cases with human effort alone. While data augmentation is an effective technique to enrich the training data, existing methods for camera data in autonomous driving applications are confined to the 2D image plane, which may not optimally increase data diversity in 3D real-world scenarios. To this end, we propose a 3D data augmentation approach termed Drive-3DAug, aiming at augmenting the driving scenes on camera in the 3D space. We first utilize Neural Radiance Field (NeRF) to reconstruct the 3D models of background and foreground objects. Then, augmented driving scenes can be obtained by placing the 3D objects with adapted location and orientation at the pre-defined valid region of backgrounds. As such, the training database could be effectively scaled up. However, the 3D object modeling is constrained to the image quality and the limited viewpoints. To overcome these problems, we modify the original NeRF by introducing a geometric rectified loss and a symmetric-aware training strategy. We evaluate our method for the camera-only monocular 3D detection task on the Waymo and nuScences datasets. The proposed data augmentation approach contributes to a gain of 1.7% and 1.4% in terms of detection accuracy, on Waymo and nuScences respectively. Furthermore, the constructed 3D models serve as digital driving assets and could be recycled for different detectors or other 3D perception tasks.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

3D Data Augmentation for Driving Scenes on Camera

Tong¹,

Xie²,

Li³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…In contrast to implicit rendering, point cloud rendering [1,6,13,18,33,36,59] is a promising editable rendering model. On the one hand, explicit 3D representations are better for interactive editing.…”

Section: Introductionmentioning

confidence: 99%