Method

MonoCD + DG+WA [mo] [lp] [MonoCD+DGWA]


Submitted on 17 Mar. 2026 06:57 by
Taewook Eum (SKKU)

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

Method Description:
Post-refinement module for monocular 3D
detection. Uses MonoCD as the base
detector with sparse LiDAR for geometric
correction. Visible-face reasoning
identifies which bounding-box surfaces face the
camera, enabling surface-to-center
offset correction. A weighted anchor mechanism
aggregates LiDAR points using
depth, face proximity, density, and range weights.
A dynamic utility gate
selectively applies correction only when LiDAR
evidence is reliable.
Parameters:
\tau=0.25, \sigma_d=max(0.15*z_{mono}, 0.5),
\sigma_f=0.4*max(w,l), r_{density}=0.3m
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) 84.18 % 61.65 % 49.38 %
Car (Orientation) 84.07 % 61.51 % 49.25 %
Car (3D Detection) 15.85 % 10.35 % 7.99 %
Car (Bird's Eye View) 27.81 % 20.06 % 14.60 %
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