Method

Fusion between PC_CNN and point cloud [la] [F-PC_CNN]
[Anonymous Submission]

Submitted on 19 Sep. 2017 20:18 by
[Anonymous Submission]

Running time:0.5 s
Environment:GPU @ 3.0 Ghz (Matlab + C/C++)

Method Description:
We designed a general pipeline for 3D vehicle detection. It
fuses the existing 2D detection network with a point cloud to
output 3D information.
Parameters:
Latex Bibtex:

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) 59.99 % 47.78 % 45.83 %
Car (3D Detection) 50.46 % 42.67 % 40.15 %
Car (Bird's Eye View) 77.05 % 69.77 % 62.59 %
This table as LaTeX


2D object detection 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