Method

Instance Centric Densification via Panoptic Segmentation and Interpolation for Object Dectection [ICD-PSI]


Submitted on 16 Oct. 2025 01:34 by
Yahya Tauk (Université de Technologie de Compiègne)

Running time:0.1 s
Environment:1 core @ 2.5 Ghz (C/C++)

Method Description:
In this paper we propose a pipeline called Instance
Centric Densification via LiDAR Panoptic
Segmentation for Object Dectection. First, we
extract instance masks via panoptic segmentation. We
then densify only OOIs instances. Thereafter, we
reinject those densified OOIs into the original
LiDAR scene and then do object detection.
Parameters:
TBD
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) 98.09 % 97.73 % 92.94 %
Car (Orientation) 98.08 % 97.59 % 92.73 %
Car (3D Detection) 90.55 % 84.35 % 79.61 %
Car (Bird's Eye View) 94.60 % 91.29 % 88.48 %
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