Method

keras-yolo3 [Kyolo3]
[Anonymous Submission]

Submitted on 17 Jun. 2018 06:08 by
[Anonymous Submission]

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

Method Description:
yolov3 without pretrain 45epoch
Parameters:
learning rate=0.0001
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) 47.18 % 33.01 % 27.57 %
Car (Orientation) 19.50 % 18.21 % 15.99 %
Pedestrian (Detection) 25.73 % 20.99 % 20.51 %
Pedestrian (Orientation) 12.06 % 9.67 % 9.27 %
Cyclist (Detection) 9.09 % 9.09 % 9.09 %
Cyclist (Orientation) 3.96 % 3.96 % 3.96 %
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