Method

ZKNet [ZKNet]


Submitted on 3 Dec. 2018 10:17 by
luo zekun (umich)

Running time:0.01 s
Environment:GPU @ 2.0 Ghz (Python)

Method Description:
A fast object detection method for autonomous driving
Parameters:
image size is 1280*384
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) 92.17 % 82.96 % 72.43 %
Car (Orientation) 38.09 % 34.27 % 29.93 %
Pedestrian (Detection) 71.15 % 56.58 % 51.87 %
Pedestrian (Orientation) 39.55 % 31.21 % 28.61 %
Cyclist (Detection) 66.29 % 49.48 % 42.81 %
Cyclist (Orientation) 28.26 % 21.51 % 18.83 %
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