Method

Subdivision Coding Network for Object Detection [la] [SCNet]


Submitted on 10 Sep. 2019 15:57 by
Li Wang (National University of Defense Technology)

Running time:0.04 s
Environment:GPU @ 3.0 Ghz (Python)

Method Description:
SCNet divides each grid into smaller sub-grids to preserve more point cloud information and converts points in the
grid to a uniform feature representation through 2D convolutional neural networks.
Parameters:
Latex Bibtex:
@ARTICLE{8813061,
author={Z. {Wang} and H. {Fu} and L. {Wang} and L. {Xiao} and B. {Dai}},
journal={IEEE Access},
title={SCNet: Subdivision Coding Network for Object Detection Based on 3D Point Cloud},
year={2019},
volume={7},
number={},
pages={120449-120462},
keywords={Three-dimensional displays;Feature extraction;Two dimensional displays;Object
detection;Proposals;Laser radar;Convolution;Autonomous driving;3D object detection;point rearrangement;sub-
grid},
doi={10.1109/ACCESS.2019.2937676},
ISSN={2169-3536},
month={},}
publisher={IEEE}
}

Detailed Results

Object detection and orientation estimation results. Results for object detection are given in terms of average precision (AP) and results for joint object detection and orientation estimation are provided in terms of average orientation similarity (AOS).


Benchmark Easy Moderate Hard
Car (Detection) 95.59 % 90.30 % 85.09 %
Car (Orientation) 95.23 % 89.36 % 84.03 %
Car (3D Detection) 83.34 % 73.17 % 67.93 %
Car (Bird's Eye View) 90.07 % 86.48 % 81.30 %
Pedestrian (Detection) 60.95 % 49.61 % 46.91 %
Pedestrian (Orientation) 44.50 % 35.49 % 33.38 %
Pedestrian (3D Detection) 47.83 % 38.66 % 35.70 %
Pedestrian (Bird's Eye View) 56.87 % 46.73 % 42.74 %
Cyclist (Detection) 78.48 % 62.50 % 56.34 %
Cyclist (Orientation) 77.77 % 61.11 % 54.82 %
Cyclist (3D Detection) 67.98 % 50.79 % 45.15 %
Cyclist (Bird's Eye View) 73.73 % 56.39 % 49.99 %
This table as LaTeX


2D object detection results.
This figure as: png eps pdf txt gnuplot



Orientation estimation results.
This figure as: png eps pdf txt gnuplot



3D object detection results.
This figure as: png eps pdf txt gnuplot



Bird's eye view results.
This figure as: png eps pdf txt gnuplot



2D object detection results.
This figure as: png eps pdf txt gnuplot



Orientation estimation results.
This figure as: png eps pdf txt gnuplot



3D object detection results.
This figure as: png eps pdf txt gnuplot



Bird's eye view results.
This figure as: png eps pdf txt gnuplot



2D object detection results.
This figure as: png eps pdf txt gnuplot



Orientation estimation results.
This figure as: png eps pdf txt gnuplot



3D object detection results.
This figure as: png eps pdf txt gnuplot



Bird's eye view results.
This figure as: png eps pdf txt gnuplot




eXTReMe Tracker