Method

soft-retinanet [softretina]


Submitted on 26 Aug. 2018 04:40 by
yu yijie (Jiangsu University)

Running time:0.16 s
Environment:4 cores @ 2.5 Ghz (Python)

Method Description:
resnet101+softnms
Parameters:
100epoch
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) 89.36 % 79.15 % 69.24 %
Car (Orientation) 37.63 % 32.90 % 28.73 %
Pedestrian (Detection) 0.68 % 0.93 % 0.95 %
Pedestrian (Orientation) 0.35 % 0.49 % 0.50 %
Cyclist (Detection) 0.29 % 0.44 % 0.22 %
Cyclist (Orientation) 0.14 % 0.20 % 0.10 %
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