Method

MonoCtrl: Controlled Cue Mining and Enhancement for Monocular 3D Object Detection [MonoCtrl_MonoDistill]
[Anonymous Submission]

Submitted on 23 Oct. 2025 17:03 by
[Anonymous Submission]

Running time:0.06 s
Environment:1 core @ 2.5 Ghz (Python)

Method Description:
This paradigm mines discriminative cues directly
from monocular images, enabling targeted
enhancement of detection performance for occluded
and distant objects. Furthermore, it is designed
to seamlessly integrate into existing monocular
detection frameworks without extensive
modifications.
Parameters:
\alpha=1
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) 77.75 % 81.47 % 76.88 %
Car (Orientation) 77.22 % 80.34 % 75.13 %
Car (3D Detection) 22.16 % 18.24 % 16.63 %
Car (Bird's Eye View) 30.36 % 25.58 % 23.60 %
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