Method

Edge-based Monocular 3D Object Detection [MonoEdge-Rotate]
[Anonymous Submission]

Submitted on 18 Nov. 2021 18:27 by
[Anonymous Submission]

Running time:0.05 s
Environment:GPU @ 2.5 Ghz (Python)

Method Description:
We leverage the local perspectives to locate objects
in 3D from a single image by regressing key-edges
with ROI-rotations.
Parameters:
n/a
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) 93.53 % 90.30 % 80.60 %
Car (Orientation) 93.46 % 90.07 % 80.29 %
Car (3D Detection) 21.60 % 14.13 % 12.27 %
Car (Bird's Eye View) 31.19 % 20.16 % 17.35 %
Pedestrian (Detection) 62.79 % 47.56 % 43.02 %
Pedestrian (Orientation) 58.58 % 43.68 % 39.34 %
Pedestrian (3D Detection) 10.47 % 7.02 % 5.84 %
Pedestrian (Bird's Eye View) 11.62 % 7.53 % 6.79 %
Cyclist (Detection) 68.91 % 51.89 % 45.33 %
Cyclist (Orientation) 61.47 % 44.56 % 38.87 %
Cyclist (3D Detection) 4.28 % 2.51 % 2.13 %
Cyclist (Bird's Eye View) 5.36 % 3.01 % 2.83 %
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



3D object detection results.
This figure as: png eps pdf txt gnuplot



Bird's eye view 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



3D object detection results.
This figure as: png eps pdf txt gnuplot



Bird's eye view 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



3D object detection results.
This figure as: png eps pdf txt gnuplot



Bird's eye view results.
This figure as: png eps pdf txt gnuplot




eXTReMe Tracker