Method

cascade_merge [Cmerge]


Submitted on 5 Oct. 2018 07:01 by
yu yijie (Jiangsu University)

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

Method Description:
cascade merge retinanet
Parameters:
1e-4 0.85weight cascade +0.6 retinanet
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) 94.75 % 86.52 % 70.20 %
Car (Orientation) 40.85 % 37.17 % 30.53 %
Pedestrian (Detection) 68.13 % 51.52 % 50.56 %
Pedestrian (Orientation) 37.14 % 28.05 % 27.43 %
Cyclist (Detection) 63.25 % 50.65 % 44.64 %
Cyclist (Orientation) 26.30 % 21.62 % 18.93 %
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