Method

centernet_baseline_resnet18 [centernet]


Submitted on 15 Jul. 2019 07:04 by
Hantao Wang (Fuzhou University)

Running time:0.01 s
Environment:GPU @ 2.5 Ghz (Python)

Method Description:
The baseline of centernet
Parameters:
backbone: resnet18
width=1280
height=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) 89.78 % 79.97 % 70.57 %
Car (Orientation) 37.83 % 34.18 % 30.75 %
Pedestrian (Detection) 68.65 % 51.75 % 46.89 %
Pedestrian (Orientation) 40.82 % 31.62 % 28.94 %
Cyclist (Detection) 59.28 % 44.50 % 39.75 %
Cyclist (Orientation) 23.40 % 18.36 % 16.35 %
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