Method

HM-SSD-3DV3 [HM3D]


Submitted on 27 May. 2017 04:02 by
Xiaofei Wu (Univ of Hong Kong)

Running time:0.35 s
Environment:GPU @ >3.5 Ghz (C/C++)

Method Description:
ssd 2d detection + orientation estimation subnet
Parameters:
\alpha = 0.2
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) 89.86 % 88.26 % 78.24 %
Car (Orientation) 89.41 % 87.29 % 77.08 %
Pedestrian (Detection) 77.60 % 65.97 % 61.09 %
Pedestrian (Orientation) 70.22 % 58.21 % 53.72 %
Cyclist (Detection) 76.28 % 63.89 % 56.51 %
Cyclist (Orientation) 67.32 % 55.12 % 48.86 %
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



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



Orientation estimation 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




eXTReMe Tracker