Method

Object-Aware Centroid Voting for Monocular 3D Object Detection [OACV]
[Anonymous Submission]

Submitted on 7 Nov. 2019 20:32 by
[Anonymous Submission]

Running time:0.23 s
Environment:GPU @ 2.5 Ghz (Python)

Method Description:
Based on the observations of apparent height
invariance in autonomous driving scenario, the
grid coordinates of 2D box are used to estimate
the 3D centroids.Besides, by leveraging the
region-wise appearance attention and the
projection offset prior, a novel object-aware
voting approach is proposed to obtain the 3D
object location.With the late fusion and the
predicted 3D orientation and dimension, the 3D
amodal bounding boxes can be detected from a
single RGB image.
Parameters:
To appear in paper.
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) 94.23 % 91.21 % 83.07 %
Car (Orientation) 93.95 % 90.35 % 81.90 %
Car (3D Detection) 8.13 % 4.77 % 3.78 %
Car (Bird's Eye View) 16.24 % 10.13 % 8.28 %
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




eXTReMe Tracker