Method

LiDAR-based 3D Object Detection with Key Point Transformer [KPTr]


Submitted on 29 Mar. 2024 13:45 by
Jie Cao (西华大学)

Running time:0.07 s
Environment:1 core @ 2.5 Ghz (C/C++)

Method Description:
Our approach applies transformer to the critical
point level. Meanwhile, feature extraction in
pointnet++ is improved.
Parameters:
a=1
Latex Bibtex:
LiDAR-based 3D Object Detection with Key Point
Transformer

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) 96.55 % 93.73 % 90.84 %
Car (Orientation) 96.52 % 93.59 % 90.58 %
Car (3D Detection) 88.52 % 80.40 % 75.28 %
Car (Bird's Eye View) 92.74 % 89.48 % 84.50 %
Pedestrian (Detection) 69.70 % 59.79 % 56.03 %
Pedestrian (Orientation) 66.74 % 56.50 % 52.72 %
Pedestrian (3D Detection) 48.85 % 42.76 % 39.49 %
Pedestrian (Bird's Eye View) 53.98 % 46.83 % 44.56 %
Cyclist (Detection) 87.31 % 79.55 % 72.47 %
Cyclist (Orientation) 87.02 % 79.10 % 72.03 %
Cyclist (3D Detection) 80.73 % 67.29 % 60.38 %
Cyclist (Bird's Eye View) 83.83 % 72.24 % 63.94 %
This table as LaTeX


2D object detection results.
This figure as: png eps txt gnuplot



Orientation estimation results.
This figure as: png eps txt gnuplot



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



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



2D object detection results.
This figure as: png eps txt gnuplot



Orientation estimation results.
This figure as: png eps txt gnuplot



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



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



2D object detection results.
This figure as: png eps txt gnuplot



Orientation estimation results.
This figure as: png eps txt gnuplot



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



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




eXTReMe Tracker