Method

Corner Uncertainty PointRCNN [CU-PointRCNN]


Submitted on 20 Mar. 2020 19:08 by
Hujie Pan (Shanghai Jiao Tong University; University of California, Berkeley)

Running time:0.1 s
Environment:GPU @ 1.5 Ghz (Python + C/C++)

Method Description:
2-Stage Method with corner transformation and
uncertainty modeling
Parameters:
See in the paper later
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) 97.25 % 91.34 % 86.98 %
Car (Orientation) 97.24 % 91.25 % 86.85 %
Car (3D Detection) 86.55 % 76.87 % 73.17 %
Car (Bird's Eye View) 92.65 % 86.69 % 82.66 %
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