Method

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

Submitted on 11 Jan. 2018 05:48 by
[Anonymous Submission]

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

Method Description:
This work is based on our previous method F-PC_CNN.
Further fusion between multiple views through deep
learning framework is added to the previous work to
improve the results.
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) 90.78 % 89.45 % 82.77 %
Car (3D Detection) 80.28 % 66.22 % 58.94 %
Car (Bird's Eye View) 88.89 % 80.69 % 76.04 %
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