Method

Frustum 3DNet [F-3DNet]


Submitted on 18 Mar. 2020 07:39 by
Fenglei Xu (Nanjing University of Science and Technology)

Running time:0.5 s
Environment:GPU @ 2.5 Ghz (Python)

Method Description:
F-3DNet: Extracting Inner order of Point Cloud
for 3D Object Detection.
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) 96.51 % 93.38 % 88.32 %
Car (Orientation) 38.58 % 37.18 % 36.44 %
Car (3D Detection) 85.48 % 78.48 % 71.62 %
Car (Bird's Eye View) 92.68 % 88.76 % 83.63 %
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