While radar and video data can be readily fused at the detection level,
fusing them at the pixel level is potentially more beneficial. This is also
more challenging in part due to the sparsity of radar, but also because
automotive radar beams are much wider than a typical pixel combined with a
large baseline between camera and radar, which results in poor association
between radar pixels and color pixel. A consequence is that depth completion
methods designed for LiDAR and video fare poorly for radar and video. Here we
propose a radar-to-pixel association stage which learns a mapping from radar
returns to pixels. This mapping also serves to densify radar returns. Using
this as a first stage, followed by a more traditional depth completion method,
we are able to achieve image-guided depth completion with radar and video. We
demonstrate performance superior to camera and radar alone on the nuScenes
dataset. Our source code is available at https://github.com/longyunf/rc-pda.
Description
Radar-Camera Pixel Depth Association for Depth Completion
%0 Conference Paper
%1 2021-long
%A Long, Yunfei
%A Morris, Daniel
%A Liu, Xiaoming
%A Castro, Marcos
%A Chakravarty, Punarjay
%A Narayanan, Praveen
%B IEEE Conference on Computer Vision and Pattern Recognition, 2021
%D 2021
%K association camera completion depth monocular nuscenes pixel radar
%T Radar-Camera Pixel Depth Association for Depth Completion
%U http://arxiv.org/abs/2106.02778
%X While radar and video data can be readily fused at the detection level,
fusing them at the pixel level is potentially more beneficial. This is also
more challenging in part due to the sparsity of radar, but also because
automotive radar beams are much wider than a typical pixel combined with a
large baseline between camera and radar, which results in poor association
between radar pixels and color pixel. A consequence is that depth completion
methods designed for LiDAR and video fare poorly for radar and video. Here we
propose a radar-to-pixel association stage which learns a mapping from radar
returns to pixels. This mapping also serves to densify radar returns. Using
this as a first stage, followed by a more traditional depth completion method,
we are able to achieve image-guided depth completion with radar and video. We
demonstrate performance superior to camera and radar alone on the nuScenes
dataset. Our source code is available at https://github.com/longyunf/rc-pda.
@inproceedings{2021-long,
abstract = {While radar and video data can be readily fused at the detection level,
fusing them at the pixel level is potentially more beneficial. This is also
more challenging in part due to the sparsity of radar, but also because
automotive radar beams are much wider than a typical pixel combined with a
large baseline between camera and radar, which results in poor association
between radar pixels and color pixel. A consequence is that depth completion
methods designed for LiDAR and video fare poorly for radar and video. Here we
propose a radar-to-pixel association stage which learns a mapping from radar
returns to pixels. This mapping also serves to densify radar returns. Using
this as a first stage, followed by a more traditional depth completion method,
we are able to achieve image-guided depth completion with radar and video. We
demonstrate performance superior to camera and radar alone on the nuScenes
dataset. Our source code is available at https://github.com/longyunf/rc-pda.},
added-at = {2021-07-05T22:21:38.000+0200},
author = {Long, Yunfei and Morris, Daniel and Liu, Xiaoming and Castro, Marcos and Chakravarty, Punarjay and Narayanan, Praveen},
biburl = {https://www.bibsonomy.org/bibtex/2b77fb49f685ccea9c8c0c0d3e11c22cb/pkoch},
booktitle = {IEEE Conference on Computer Vision and Pattern Recognition, 2021},
description = {Radar-Camera Pixel Depth Association for Depth Completion},
interhash = {3a30c469fef826b1d297896d9c1f9e4c},
intrahash = {b77fb49f685ccea9c8c0c0d3e11c22cb},
keywords = {association camera completion depth monocular nuscenes pixel radar},
timestamp = {2021-07-15T14:01:29.000+0200},
title = {Radar-Camera Pixel Depth Association for Depth Completion},
url = {http://arxiv.org/abs/2106.02778},
year = 2021
}