Method

Next-generation Accurate and Efficient Monocular 3D Object Detection Method Fast [monodetrnext-f]
[Anonymous Submission]

Submitted on 8 May. 2024 13:26 by
[Anonymous Submission]

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

Method Description:
Informed by efficacious strategies within the
domain of 2D detection, we introduce MonoDETRNext
with the explicit goal of achieving a harmonious
equilibrium between precision and speed to
surmount these inherent challenges. Building upon
the foundational principles of MonoDETR, we
delineate two distinct variants: MonoDETRNext-F,
prioritizing velocity, and MonoDETRNext-A,
emphasizing precision-centric objectives.
Parameters:
~40M
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) 89.19 % 76.64 % 69.75 %
Car (Orientation) 88.71 % 75.59 % 68.62 %
Car (3D Detection) 27.21 % 21.69 % 21.16 %
Car (Bird's Eye View) 34.56 % 28.12 % 28.33 %
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




eXTReMe Tracker