Method

Monocular 3D Object Detection: An Extrinsic Parameter Free Approach [MonoEF]


Submitted on 14 Nov. 2020 08:49 by
yunsong zhou (Shanghai Jiao Tong University)

Running time:0.03 s
Environment:1 core @ 2.5 Ghz (Python)

Method Description:
we propose a novel method to capture camera poses
to protect the model free from camera extrinsic
perturbations and improve accuracy in real-life
scenarios. Specifically, the proposed detector
predicts camera extrinsic parameters by detecting
vanishing point and horizon changes and and
converter uses them to correct for perturbed
image features in latent space. Finally, the 3D
detector works independent of extrinsic parameter
variations, producing accurate results even on
potholed and uneven roads.
Parameters:
None
Latex Bibtex:
@InProceedings{Zhou_2021_CVPR,
author = {Zhou, Yunsong and He, Yuan and
Zhu, Hongzi and Wang, Cheng and Li, Hongyang and
Jiang, Qinhong},
title = {Monocular 3D Object Detection: An
Extrinsic Parameter Free Approach},
booktitle = {Proceedings of the IEEE/CVF
Conference on Computer Vision and Pattern
Recognition (CVPR)},
month = {June},
year = {2021},
pages = {7556-7566}
}

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) 96.32 % 90.88 % 83.27 %
Car (Orientation) 96.19 % 90.65 % 82.95 %
Car (3D Detection) 21.29 % 13.87 % 11.71 %
Car (Bird's Eye View) 29.03 % 19.70 % 17.26 %
Pedestrian (Detection) 58.79 % 43.73 % 39.45 %
Pedestrian (Orientation) 47.45 % 34.63 % 31.01 %
Pedestrian (3D Detection) 4.27 % 2.79 % 2.21 %
Pedestrian (Bird's Eye View) 4.61 % 3.05 % 2.85 %
Cyclist (Detection) 51.06 % 41.19 % 35.70 %
Cyclist (Orientation) 43.70 % 32.19 % 27.93 %
Cyclist (3D Detection) 1.80 % 0.92 % 0.71 %
Cyclist (Bird's Eye View) 2.36 % 1.18 % 1.15 %
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



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



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