Method

Associate-3Ddet: Perceptual-to-Conceptual association for 3D Point Cloud Object Detection [Associate-3Ddet]
https://github.com/dleam/Associate-3Ddet

Submitted on 19 Nov. 2019 05:01 by
Liang Du (Fudan University)

Running time:0.05 s
Environment:1 core @ 2.5 Ghz (Python + C/C++)

Method Description:
We introduce a brain-inspired 3D object detection
framework, which mimics the functionality of the
human brain when proceeding object perception and
fundamentally boosts the performance of the 3D
object detector. In addition, given a 3D object
detection method, our approach enhances its
feature extraction ability during the training
process without introducing any extra component
in inference stage, which makes our framework is
easy to integrate into many 3D object detection
methods.
Parameters:
TBD
Latex Bibtex:
@InProceedings{Du_2020_CVPR,
author = {Du, Liang and Ye, Xiaoqing and Tan, Xiao
and Feng, Jianfeng and Xu, Zhenbo and Ding, Errui
and Wen, Shilei},
title = {Associate-3Ddet: Perceptual-to-Conceptual
Association for 3D Point Cloud Object Detection},
booktitle = {The IEEE/CVF Conference on Computer
Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2020}
}

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) 95.61 % 92.45 % 87.32 %
Car (Orientation) 0.52 % 1.20 % 1.38 %
Car (3D Detection) 85.99 % 77.40 % 70.53 %
Car (Bird's Eye View) 91.40 % 88.09 % 82.96 %
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