Method

Retinanet100epoch [Retinanet100]


Submitted on 11 Jul. 2018 05:01 by
Yue Li (Southern Methodist University)

Running time:0.2 s
Environment:4 cores @ 2.5 Ghz (Python)

Method Description:
100epoch to write
Parameters:
1e-4 learning rate
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) 89.83 % 78.85 % 68.73 %
Car (Orientation) 37.54 % 32.87 % 28.69 %
Pedestrian (Detection) 52.43 % 42.83 % 35.02 %
Pedestrian (Orientation) 28.72 % 23.23 % 19.00 %
Cyclist (Detection) 46.39 % 37.54 % 30.82 %
Cyclist (Orientation) 18.64 % 15.16 % 12.49 %
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