CaltechAUTHORS
  A Caltech Library Service

VoxFormer: Sparse Voxel Transformer for Camera-based 3D Semantic Scene Completion

Li, Yiming and Yu, Zhiding and Choy, Christopher and Xiao, Chaowei and Alvarez, Jose M. and Fidler, Sanja and Feng, Chen and Anandkumar, Anima (2023) VoxFormer: Sparse Voxel Transformer for Camera-based 3D Semantic Scene Completion. . (Unpublished) https://resolver.caltech.edu/CaltechAUTHORS:20230316-183809946

[img] PDF - Submitted Version
See Usage Policy.

4MB

Use this Persistent URL to link to this item: https://resolver.caltech.edu/CaltechAUTHORS:20230316-183809946

Abstract

Humans can easily imagine the complete 3D geometry of occluded objects and scenes. This appealing ability is vital for recognition and understanding. To enable such capability in AI systems, we propose VoxFormer, a Transformer-based semantic scene completion framework that can output complete 3D volumetric semantics from only 2D images. Our framework adopts a two-stage design where we start from a sparse set of visible and occupied voxel queries from depth estimation, followed by a densification stage that generates dense 3D voxels from the sparse ones. A key idea of this design is that the visual features on 2D images correspond only to the visible scene structures rather than the occluded or empty spaces. Therefore, starting with the featurization and prediction of the visible structures is more reliable. Once we obtain the set of sparse queries, we apply a masked autoencoder design to propagate the information to all the voxels by self-attention. Experiments on SemanticKITTI show that VoxFormer outperforms the state of the art with a relative improvement of 20.0% in geometry and 18.1% in semantics and reduces GPU memory during training by ~45% to less than 16GB. Our code is available on this https://github.com/NVlabs/VoxFormer.


Item Type:Report or Paper (Discussion Paper)
Related URLs:
URLURL TypeDescription
http://arxiv.org/abs/2302.12251arXivDiscussion Paper
https://github.com/NVlabs/VoxFormerRelated ItemCode
ORCID:
AuthorORCID
Li, Yiming0000-0002-0157-6218
Yu, Zhiding0000-0003-1776-996X
Choy, Christopher0000-0002-6566-3193
Xiao, Chaowei0000-0002-7043-4926
Fidler, Sanja0000-0003-1040-3260
Feng, Chen0000-0003-3211-1576
Anandkumar, Anima0000-0002-6974-6797
Record Number:CaltechAUTHORS:20230316-183809946
Persistent URL:https://resolver.caltech.edu/CaltechAUTHORS:20230316-183809946
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:120092
Collection:CaltechAUTHORS
Deposited By: George Porter
Deposited On:16 Mar 2023 22:50
Last Modified:16 Mar 2023 22:50

Repository Staff Only: item control page