Method

CLOCs [CLOCs_PointCas]


Submitted on 14 Nov. 2019 05:22 by
Su Pang (MIchigan State University)

Running time:0.1 s
Environment:GPU @ 2.5 Ghz (Python)

Method Description:
A new fast fusion network to easily upgrade
performance
of single-modality detectors. This result is based
on fusing PointRCNN and CascadeRCNN.
Parameters:
TBD
Latex Bibtex:
TBD

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) 96.69 % 93.55 % 86.16 %
Car (Orientation) 96.66 % 93.34 % 85.87 %
Car (3D Detection) 87.50 % 76.68 % 71.21 %
Car (Bird's Eye View) 92.60 % 88.99 % 81.74 %
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