Method

yolo4_5l [yolo4_5l]
https://edu.icoremail.net/coremail/XT5/index.jsp?sid=BAduTzoooVfEXWGShQoovYaMIaopYYgR#mail.list%7C%7

Submitted on 16 Jun. 2020 10:48 by
Hai Wang (Jiangsu University)

Running time:0.02 s
Environment:1 core @ 2.5 Ghz (Python + C/C++)

Method Description:
add two yolo layers to improve the ablity of detect
on small target.
Parameters:
q = 1
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) 91.79 % 90.38 % 80.64 %
Car (Orientation) 37.14 % 36.81 % 33.24 %
Pedestrian (Detection) 71.89 % 52.74 % 47.90 %
Pedestrian (Orientation) 38.95 % 28.60 % 25.97 %
Cyclist (Detection) 69.14 % 48.38 % 42.16 %
Cyclist (Orientation) 28.67 % 20.79 % 18.35 %
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